Abstract

We experimentally demonstrate a multi-task deep neutral network (MT-DNN) enabled optical performance monitoring (OPM) for PDM-QPSK/8QAM/16QAM signals, by using the asynchronous amplitude histogram (AAH) after the direct detection. Consequently, we can simultaneously realize the modulation format identification (MFI), baud rate identification (BRI), and optical signal-to-noise ratio (OSNR) monitoring simultaneously. In the simulation, when both 20Gbaud and 30Gbaud PDM-QPSK, PDM-8QAM and PDM-16QAM signals are taken into account, the accuracies of both MFI and BRI are 100%. Meanwhile, the root-mean-square error (RMSE) of OSNR monitoring is 0.58dB over a range of 10-22dB, 14-24dB, and 17-26dB for PDM-QPSK, PDM-8QAM and PDM-16QAM, respectively. Furthermore, the numerical results show that RMSE of OSNR monitoring is degraded from 0.58dB to 0.97dB, when chromatic dispersion (CD) is accumulated from 0 to 1600ps/nm. Meanwhile, the MFI accuracy is degraded from 100% to 97.25%, and the BRI accuracy remains 100%. When 2.8Gbaud and 9.8Gbaud signals are used for the experimental verification under the condition of back-to-back transmission, the accuracies of MFI and BRI are 100% and 96.8%, respectively, and the RMSE of OSNR monitoring is 0.76dB. Since only one photodetector (PD) with the asynchronous sampling and one MT-DNN are required, the proposed OPM scheme has the advantage of high cost performance.

© 2019 Optical Society of America under the terms of the OSA Open Access Publishing Agreement

1. Introduction

With the explosion of big data demand, the capacity of optical network has kept increasing rapidly and the network architectures continuously become more dynamic, complex, and transparent. Therefore, optical performance monitoring (OPM) is of great significance to ensure the reliable operation of optical network [1]. As for various monitoring parameters, optical signal-to-noise ratio (OSNR), because of its straight-forward relationship to the bit error ratio (BER) of transmission system, is primarily desired [2]. So far, there exist many traditional OSNR monitoring techniques for coherent fiber optical transmission systems. Some techniques require additional photonic components such as interferometers [3] or optical filters [4], indicating of both high implementation complexity and modulation format sensitive operation. As for the error vector magnitude (EVM) based technique [5], coherent receive is required to realize the OSNR monitoring. For the technique based on Golay sequences [6], additional data sequences are compulsory for the OSNR monitoring, leading to the spectral efficiency (SE) penalty of fiber optical transmission. For the polarization nulling based OSNR monitoring method [7], it cannot support the polarization division multiplexing (PDM) technique for the modulation format of signal is limited to non-return-to-zero (NRZ) signal.

Due to the development of elastic optical network, the standard single mode fiber (SSMF) link condition of optical network becomes more dynamic [8]. Therefore, the selected modulation format and baud rate are closely related to the digital signal processing (DSP) algorithm at the receiver side, by taking the overall SSMF link condition into account. As a result, both modulation formation identification (MFI) and baud rate identification (BRI) are important parameters to be monitored during the OPM implementation. So far, existing MFI and BRI monitoring techniques include amplitude histograms (AHs) [9,10], asynchronous delay-tap sampling portraits (ADTPs) [11,12], variational Bayesian expectation maximization algorithm [13], signal cumulants [14], and digital frequency-offset loading [15]. Many of these techniques require additional hardware components [11,12] and channel information [13–15], which can significantly increase implementation complexity and cost. Meanwhile, most of these techniques are focus on MFI without the capability of OSNR monitoring and BRI [9–15].

Recently, artificial intelligence (AI) technique has been verified as a powerful tool for various applications. Compared with the traditional techniques, the AI technique has the advantages of autonomous feature extraction and mapping, intelligent response and less manual intervention [16]. There are some research results of OPM implementation based on artificial neural network (ANN) [17–24]. However, these methods more or less have the following disadvantages: (1) some techniques require multiple ANNs or other machine learning algorithms to separately realize MFI and OSNR monitoring, leading to high calculation complexity and slow response [17–19]. (2) A convolution neural network (CNN) based technique is proposed to turn the OSNR monitoring into a classification problem [20]. In order to further improve the accuracy of OSNR monitoring, the classification interval must be small enough, resulting in a very huge amount of data sets. (3) Most of these techniques require the use of coherent receiver with high sampling rate during the OPM implementation, leading to a high cost [17,19–24]. However, for the advanced modulation formats, it is not practical to use the coherent receivers everywhere in the fiber optic link to realize the OPM. Therefore, relying on the signal amplitude information to implement the OPM with only one PD for the advanced modulation formats is ideally desired in view of practical OPM applications.

In current submission, based on a single multi-task deep neural network (MT-DNN), we can simultaneously realize MFI, BRI and OSNR monitoring, with great reduction of ANN complexity and training cost. Here, we use a MT-DNN trained with the asynchronous amplitude histogram (AAH) generated by the single PD direct detection. Once the training process of MT-DNN is finished, only single AAH is desired to be fed into the MT-DNN for the ease of OPM implementation. We experimentally verify that the parameters of three commonly-used modulation formats for current fiber optical coherent transmission, including PDM-QPSK, PDM-8QAM and PDM-16QAM signals, can be monitored without crosstalk. The MFI and BRI accuracies of three signals can reach 100% and 96.8%, respectively. Meanwhile, the RMSE of OSNR monitoring is 0.76dB over a range from 10 to 22dB, 14-24dB and 17-26dB for PDM-QPSK, PDM-8QAM and PDM-16QAM, respectively.

2. Operation principle

2.1 Asynchronous amplitude histogram

The asynchronous amplitude histogram (AAH) carries out statistics on the amplitude of data after the asynchronous sampling, as shown in Fig. 1, the vertical axis of AAH denotes the frequency of data appeared in each amplitude interval, and the horizontal axis of AAH denotes the number of the amplitudes equally divided. Obviously for various modulation formats with different baud rates, AAHs obtained under the condition of the same asynchronous sampling rate have obvious differences. Moreover, even when the OSNR is varied for a specific modulation format with fixed baud rate, the corresponding AAH still has slight difference. Subsequently, MT-DNN can automatically extract these features from AAHs, for the ease of OPM implementation. Please note that, although AAH is a two-dimensional image, the horizontal axis only represents the number of intervals with the equal step of amplitude. For DNN used in current submission, only one-dimensional data representing the frequency of occurrence is required, and the horizontal axis of AAH has less influence. In case the AAH graph is divided into 80 bins, then each one-dimensional vector with a size of 80x1 is introduced to the input of MT-DNN later. Therefore, in comparison with other techniques using CNN for two-dimensional image recognition [20,24], the complexity of our proposed scheme is greatly reduced. Furthermore, the proposed technique is cost-effective for using only one PD with asynchronous sampling rather than coherent receiver with high sampling rate and clock synchronization.

 figure: Fig. 1

Fig. 1 AAHs with 80 bins for signals with various OSNRs, modulation formats and baud rates after asynchronous sampling.

Download Full Size | PPT Slide | PDF

2.2 MT-DNN

MT-DNN is a kind of ANN based on the multi-task machine learning [25]. Most ANN can be used for solving two kinds of problems. One is used to solve the classification problem. In term of OPM, identification of both modulation format and baud rate belongs to such problem, even including other discrete classification problems. The other is used for the regression analysis. OSNR monitoring is a kind of such application. The typical structure of DNN for the classification is shown in Fig. 2(a), the last layer of DNN uses the Softmax function as the activation function so that the output is switched between 0 and 1. Softmax function can be expressed in Eq. (1), m is the number of neurons at the output layer. Each output vector corresponds to a specific situation. For example, “001” stands for “QPSK” signal, “010” stands for “8QAM” signal, and “100” stands for “16QAM” signal. However, as for the OSNR monitoring, the output value of the DNN in Fig. 2(b) needs to be continuous value rather than discrete value. Thus it does not belong to the same learning task as the classification problem. Previously, OSNR monitoring can be changed into a classification problem to be realized together with MFI [20]. Alternatively, MFI and OSNR monitoring can be implemented, respectively, with individual DNNs [17].

 figure: Fig. 2

Fig. 2 DNN structure for (a) classification and (b) regression.

Download Full Size | PPT Slide | PDF

Softmax(xi)=exiimexi

With the help of multi-task machine learning, MT-DNN can realize two kinds of tasks simultaneously. As shown in Fig. 3, the previous layers of MT-DNN used in current submission are the same as that of normal DNN, which is called as the shared layers. Only at the last layer, different activation functions and neurons are used to complete different tasks. Our OPM consists of three tasks, namely MFI for three modulation formats, BRI for two baud rates, and the OSNR monitoring for all involved signals. Since multiple tasks are introduced, the loss function L of the MT-DNN becomes:

L=λ1L1+λ2L2+λ3L3
where L1, L2 and L3 are loss functions of MFI, BRI and OSNR monitoring, respectively. λ1, λ2 and λ3 are the weights of the three tasks, respectively. Then, L1, L2 and L3 can be expressed as Eqs. (3) and (4):
L1=L2=1m[i=1myilogyi+(1yi)log(1yi)]
L3=1mi=1m(yiyi)2
where m is the number of samples, yiis the actual output, and yi is the predicted output of MT-DNN. L1 and L2 are cross entropy loss function and L3 is mean-square error (MSE) function [26]. By adjusting the value of λ1, λ2 and λ3, we can vary the weights of three tasks in order to optimize all monitoring results. Because of the multi-task set, the shared layers of DNN utilizes the same mapping model. Thus, multiple DNN implementations are avoided, leading to great complexity reduction of each training and testing.

 figure: Fig. 3

Fig. 3 MT-DNN structure with AAHs bins vector as input and identified OSNRs, symbol rates and modulation formats as output.

Download Full Size | PPT Slide | PDF

3. Simulation setup and results

3.1 Simulation setup

Firstly, we carry out numerical simulations based on VPItransmissionMaker to generate three widely-used optical signals (PDM-QPSK/8QAM/16QAM) modulated by a pseudo-random binary sequence (PRBS) with a length of 216 at 20Gbaud and 30Gbaud, as shown in Fig. 4. An erbium-doped fiber amplifier (EDFA) is used to add amplified spontaneous emission (ASE) noise into the optical signal, and a variable optical attenuator (VOA) is used to adjust OSNRs for PDM-QPSK/8QAM/16QAM optical signals over a range of 10−22 dB, 14−24 dB, and 17−26dB, respectively, with a step of 1dB. Then the signals is transmitted through a section of standard single-mode fiber (SSMF) and a polarization mode dispersion (PMD) emulator to vary the chromatic dispersion (CD) and differential group delay (DGD) values. After the optical to electronic conversion by a PD and the asynchronous sampling, we collect 30 AAHs of individual OSNR value under the condition of a specific modulation format with fixed baud rate. Thus, under the condition of back-to-back (B2B) transmission, the whole data sets for the MT-DNN comprise 2040 AAHs in total. Each AAH has 80 bins, and the AAHs in the data sets are divided into the training and testing sets by randomly selecting 80% and 20% of all AAHs. Next, the AAH images are simplified to a one-dimensional vector with a size of 80x1 and introduced into MT-DNN for the purpose of OPM implementation. In our submission, Keras library combined with Tensorflow backend are selected as the model of MT-DNN [27]. The MT-DNN used in the simulation has 5 layers, and the number of neurons in each layer is 80, 160, 80, 40, 3/2/1. The activation function of the first three layers is ReLU function, while the activation functions of the output layer for BRI, MFI and OSNR monitoring are Softmax, Softmax, Linear, respectively.

 figure: Fig. 4

Fig. 4 Simulation setup for joint MFI, BRI and OSNR monitoring with a single MT-DNN and directly detected PDM-QAM signals.

Download Full Size | PPT Slide | PDF

3.2 Results and discussions under the condition of B2B transmission

Since the weight of each task will affect the OPM results, we need to optimize the weight of each task in order to achieve the best monitoring results. When the weight of the corresponding MFI and BRI is set to 0.2 (λ1 = λ2 = 0.2), the variation of OSNR monitoring results and BR/MFI accuracy with respect to the weight of the OSNR monitoring task (λ3) is shown in Fig. 5. Since AAH characteristics of various modulation formats are quite different, MFI is relatively simple, and the accuracy can reach 100%. As for the other two tasks, there is an optimal weight for λ3 in order to simultaneously make the error of OSNR monitoring small and the accuracy of BRI high. λ1 and λ2 are optimized in the same way as λ3. After optimization, we finally choose the three task weights as 0.2, 0.2 and 0.03, respectively. Figure 6(a) shows the accuracy evolution of BRI and MFI during the training process. The accuracies of both BRI and MFI converge after 50 epochs. As shown in Table 1, the MFI of all three modulation formats can reach 100%. The BRI accuracy for all signals is 100%. Finally, as for the OSNR monitoring, RMSE of 0.58dB is observed in Fig. 6(b) for all testing data sets taken into consideration, and the maximum error of OSNR monitoring is less than 3dB. The RMSE is defined in Eq. (5), where yi is the real value, and yi is the predicted value from MT-DNN.

 figure: Fig. 5

Fig. 5 Variation of MFI, BRI and OSNR monitoring results with respect to λ3 during the testing process.

Download Full Size | PPT Slide | PDF

 figure: Fig. 6

Fig. 6 (a) MFI and BRI accuracy versus epoch during the training process, and (b) OSNR monitoring error for all signals with three modulation formats and two baud rates during the testing process of back-to-back transmission system.

Download Full Size | PPT Slide | PDF

Tables Icon

Table 1. MFI accuracies for three modulation formats during the testing process under the condition of back-to-back transmission

RMSE=i=1n(yiyi)2n

We further investigate the maximum number of modulation formats to be monitored by the proposed scheme. After the introduction of PDM-64QAM signal, we firstly used single-task DNN to carry out the OSNR monitoring for four kinds of signals. As shown in Fig. 7(a), the RMSE of OSNR monitoring for the first three modulation formats is still within 1dB, while the RMSE of OSNR monitoring for PDM-64QAM is increased to 1.97dB, as shown in Fig. 7(b). Thus, we decide to implement the proposed OPM scheme for PDM-QPSK/8QAM/16QAM signal in next experimental verification. We infer that, since the AAH diagram of PDM-64QAM signal is not sensitive with respect to the OSNR variation, it is challenging to be learned by the DNN. However, PDM-QPSK, PDM-8QAM and PDM-16QAM signals can be successfully monitored by the use of MT-DNN. Additionally, we investigate the effects of the baud rate difference on the BRI performance. In our simulation, the baud rate is reduced to below 10Gbaud for subsequent experimental verification. As shown in Fig. 8(b), for PDM-QPSK signals with an OSNR range from 10 to 22dB, when the baud rate of one signal is set to 2.8Gbaud, the BRI accuracy is improved with the growing of the baud rate of another signal. Therefore, if the difference of baud rate is not obvious, there will occur less different characteristics between two AAHs, which cannot be learned by MT-DNN, consequently leading to a degradation of BRI accuracy. When the baud rate is respectively set to 2.8Gbaud and 9.8Gbaud, the accuracy of BRI is shown in Fig. 8(a). The BRI accuracy of PDM-QPSK is 95.9%, meanwhile the BRI accuracy for the other two signals is 100%. Thus, the overall identification accuracy of BRI is 98.6%. Thus, we decide to implement the proposed OPM scheme for 2.8/9.8Gbaud signals in our next experimental verification.

 figure: Fig. 7

Fig. 7 (a) RMSE of DNN-based OSNR monitoring with respect to four modulation formats, and (b) OSNR monitoring error for 64QAM signals during the testing process.

Download Full Size | PPT Slide | PDF

 figure: Fig. 8

Fig. 8 (a) BRI accuracy for three modulation formats during the testing process, and (b) BRI accuracy versus symbol rate of QPSK signals during the testing process.

Download Full Size | PPT Slide | PDF

3.3 Effects of CD and PMD

We further investigate the tolerance of CD and PMD for our proposed OPM scheme. Here, we vary the SSMF length to bring a variable CD, and the PMD emulator is used to generate a various DGD. We firstly investigate effects of CD and PMD on the OSNR monitoring and BRI. When 20Gbaud and 30Gbaud PDM-16QAM signals are taken into account, the RMSE variation of OSNR monitoring and BRI accuracy with respect to the accumulated CD, from 0ps/nm to 1600ps/nm, is shown in Fig. 9(a) under the condition of DGD = 0ps. With the growing CD, the RMSE of OSNR monitoring is worsened from 0.68dB to 1.2dB for PDM-16QAM signal, while the BRI accuracy remains 100%. We infer that the AAH waveform varies significantly with respect to CD, due to the CD induced pulse broadening. However, the characteristics of AAH fluctuation with respect to the OSNR can still be recognized, as shown in Fig. 9(b). We can clearly observe that the OSNR monitoring results under the condition of large CD (1600ps/nm) are acceptable. Please note that, in order to realize the best OPM results for variable CD, MT-DNN needs to be re-trained because of the CD induced AAH waveform variation. Then, we investigate the effect of PMD on the OSNR monitoring and BRI. For 20Gbaud and 30Gbaud PDM-16QAM signals, when CD is set as 0ps/nm, Fig. 10(a) shows the variation of BRI and OSNR monitoring results with respect to DGD from 0ps to 10ps. Alternatively, the AAH waveform is insensitive to PMD, as shown in Fig. 10(b). Therefore, PMD may not lead to severe performance penalty of OPM.

 figure: Fig. 9

Fig. 9 (a) Variation of BRI and OSNR monitoring results with respect to CD for 20Gbaud and 30Gbaud PDM-16QAM signals during the testing process. (b) AAHs with 80 bins for 30Gbaud PDM-16QAM signals with various CDs (I) and OSNRs (II).

Download Full Size | PPT Slide | PDF

 figure: Fig. 10

Fig. 10 (a) Variation of BRI and OSNR monitoring results with respect to DGD for 20Gbaud and 30Gbaud PDM-16QAM signals during the testing process, and (b) AAHs for 30Gbaud PDM-16QAM with various DGDs (0ps,10ps) after asynchronous sampling (OSNR is 26dB).

Download Full Size | PPT Slide | PDF

Furthermore, we investigate effects of CD and PMD on MFI. When both 20Gbaud and 30Gbaud PDM-QPSK, PDM-8QAM and PDM-16QAM signals are taken into account, CD is set as 1600ps/nm and the DGD is set as 6ps. We first use a single-task DNN (ST-DNN) for MFI, the accuracy evolution of MFI during the training process is shown in Fig. 11(a), and the MFI accuracy during the testing process is 90.5%. Then, we use the proposed MT-DNN for BRI, MFI and OSNR monitoring simultaneously, Fig. 11(b) shows the accuracy evolution of BRI and MFI during the training process, and the accuracies of BRI and MFI during the testing process are 100% and 97.25%, respectively. As for the OSNR monitoring in Fig. 12, the RMSE is 0.97dB. We infer that the AAH waveform varies significantly with CD. However, the use of multi-task learning can effectively improve the accuracy of MFI, and the shared layers information of MT-DNN may lead to better training process with the help of the other learning tasks such as BRI and OSNR monitoring. Considering the cost performance, our proposed scheme has the potentials for fiber optical coherent transmission.

 figure: Fig. 11

Fig. 11 (a) MFI accuracy versus epoch during the training process with a ST-DNN, and (b) MFI and BRI accuracy versus epoch during the training process with a MT-DNN. (c) AAHs for signals at 1600ps/nm CD with various formats after the asynchronous sampling.

Download Full Size | PPT Slide | PDF

 figure: Fig. 12

Fig. 12 OSNR monitoring error for all signals at 1600ps/nm CD and 6ps DGD during the testing process.

Download Full Size | PPT Slide | PDF

4. Experimental setup and results

4.1 Experimental setup

The experimental setup is shown in Fig. 13. Due to the limited bandwidth of arbitrary waveform generator (AWG, Tektronix 7122C), three widely-used optical signals in coherent fiber optical transmission system (PDM-QPSK/8QAM/16QAM) at 2.8/9.8GBaud are generated. An Erbium-doped fiber amplifier (EDFA) and a variable optical attenuator (VOA) are employed to load the optical noise. Since we pay special attention to the multiple parameters monitoring performance with respect to the OSNR value, the OSNR of received QPSK, 8QAM, and 16QAM signals can be varied over a range of 10−22 dB, 14−24 dB, and 17−26 dB, respectively, with a step of ~1dB. At the receiver side, the optical signal is passed through a 0.8nm optical band pass filter (OBPF). The OSNRs of received signal can be measured by an optical spectrum analyzer (OSA) with a resolution of 0.1nm. After the direct detection by a PD with 3dB bandwidth of 10GHz, the electrical signals are asynchronously sampled by a digital sampling oscilloscope (DSO, Tektronix DPO73304D). After the asynchronous sampling, we collect 30 AAHs of individual OSNR value under the condition of a specific modulation format with the fixed baud rate. The whole data sets have 2040 AAHs in total. Since each AAH has 50 bins, the input layer of MT-DNN has 50 neurons. The AAHs are divided into the training and testing data sets by random selecting 80% and 20% of all AAHs. After being optimized, The MT-DNN used in the experiment has 5 layers, and the number of neurons at each layer is, respectively, 50, 100, 50, 30, 3/2/1. The task weights of MFI, BRI and OSNR monitoring are 0.2, 0.2, and 0.03, respectively.

 figure: Fig. 13

Fig. 13 Experimental setup of MFI, BRI and OSNR monitoring with a single MT-DNN and directly detected PDM-QAM signals.

Download Full Size | PPT Slide | PDF

4.2 Experimental results and discussions

The MFI results are shown in Table 2, the accuracy of MFI for the three modulation formats can reach 100%. As for the BRI, the results are similar to the simulation, as shown in Fig. 14(a). The BRI accuracy of PDM-QPSK is 92.4%, while the BRI accuracy of PDM-8QAM and PDM-16QAM is 100%. Thus, the overall identification accuracy of BRI is 96.8%. Figure 14(b) shows the experimental OSNR monitoring results. The RMSE is 0.76dB for all testing data sets taken into consideration and the maximum error is less than 3dB. We can conclude that the experimental results agree well with the numerical simulations. Since our proposed scheme only require single one PD without coherent detection and one MT-DNN for multi-parameters monitoring, instead of using multiple DNNs [16], our proposed OPM scheme has the advantages of low cost and low complexity, which is ideally desired for future agile optical network.

Tables Icon

Table 2. MFI accuracies for various modulation formats during the testing process in the experiment

 figure: Fig. 14

Fig. 14 (a) BRI accuracy for three modulation formats during the testing process, and (b) OSNR monitoring error of all testing data set in the experiment.

Download Full Size | PPT Slide | PDF

5. Conclusion

For three kinds of commonly-used PDM-QAM signals, we have realized optical performance monitoring including modulation format identification, baud rate identification, and OSNR monitoring, based on the asynchronous amplitude histogram together with a MT-DNN. In the simulation, when both 20Gbaud and 30Gbaud PDM-QPSK, PDM-8QAM and PDM-16QAM signals are taken into account, the accuracies of both MFI and BRI are 100%. Meanwhile, the RMSE of OSNR monitoring is 0.58dB over a range from 10 to 22dB, 14-24dB and 17-26dB for PDM-QPSK, PDM-8QAM and PDM-16QAM, respectively. Furthermore, the numerical results show that RMSE of OSNR monitoring is worsened from 0.58dB to 0.97dB, when CD is accumulated from 0 to 1600ps/nm. Meanwhile, the MFI accuracy is degraded from 100% to 97.25%, and the BRI accuracy remains 100%. When 2.8Gbaud and 9.8Gbaud signals are used for the experimental verification under the condition of B2B transmission, our results verify that the MFI and BRI accuracies of three modulation formats can respectively reach 100% and 96.8%. The RMSE of OSNR monitoring is 0.76dB. The proposed OPM scheme with the characteristic of low cost and simple implementation is a potential candidate for coherent fiber optic transmission system.

Funding

National Natural Science Foundation of China (61875061), and Key project of R&D Program of Hubei Province (2018AAA041).

References

1. F. N. Khan, Z. Dong, C. Lu, A. P. T. Lau, X. Zhou, and C. Xie, Optical performance monitoring for fiber-optic communication networks (Wiley, 2016).

2. Z. Dong, F. N. Khan, Q. Sui, K. Zhong, C. Lu, and A. P. T. Lau, “Optical performance monitoring: A review of current and future technologies,” J. Lightwave Technol. 34(2), 525–543 (2016). [CrossRef]  

3. M. R. Chitgarha, S. Khaleghi, W. Daab, A. Almaiman, M. Ziyadi, A. Mohajerin-Ariaei, D. Rogawski, M. Tur, J. D. Touch, V. Vusirikala, W. Zhao, and A. E. Willner, “Demonstration of in-service wavelength division multiplexing optical-signal-to-noise ratio performance monitoring and operating guidelines for coherent data channels with different modulation formats and various baud rates,” Opt. Lett. 39(6), 1605–1608 (2014). [CrossRef]   [PubMed]  

4. S. Oda, J. Yang, Y. Akasaka, K. Sone, Y. Aoki, M. Sekiya, and J. C. Rasmussen, “In-band OSNR monitor using an optical bandpass filter and optical power measurements for superchannel signals,” in European Conference on Optical Communication (London, U.K), 1–3 (2013). [CrossRef]  

5. R. Schmogrow, B. Nebendahl, M. Winter, A. Josten, D. Hillerkuss, S. Koenig, J. Meyer, M. Dreschmann, M. Huebner, C. Koos, J. Becker, W. Freude, and J. Leuthold, “Error vector magnitude as a performance measure for advanced modulation formats,” IEEE Photonic. Tech. L. 24(1), 61–63 (2012). [CrossRef]  

6. C. Do, A. V. Tran, C. Zhu, D. Hewitt, and E. Skafidas, “Data-aided OSNR estimation for QPSK and 16-QAM coherent optical system,” IEEE Photonics J. 5(5), 6601609 (2013). [CrossRef]  

7. J. H. Lee, D. K. Jung, C. H. Kim, and Y. C. Chung, “OSNR monitoring technique using polarization-nulling method,” IEEE Photonic. Tech. L. 13(1), 88–90 (2001). [CrossRef]  

8. O. Gerstel, M. Jinno, A. Lord, and S. B. Yoo, “Elastic optical networking: A new dawn for the optical layer?” IEEE Commun. Mag. 50(2), s12–s20 (2012). [CrossRef]  

9. F. N. Khan, Y. Zhou, A. P. T. Lau, and C. Lu, “Modulation format identification in heterogeneous fiber-optic networks using artificial neural networks,” Opt. Express 20(11), 12422–12431 (2012). [CrossRef]   [PubMed]  

10. Z. Wan, Z. Yu, L. Shu, Y. Zhao, H. Zhang, and K. Xu, “Intelligent optical performance monitor using multi-task learning based artificial neural network,” Opt. Express 27(8), 11281–11291 (2019). [CrossRef]   [PubMed]  

11. F. N. Khan, Y. Yu, M. C. Tan, W. H. Al-Arashi, C. Yu, A. P. T. Lau, and C. Lu, “Experimental demonstration of joint OSNR monitoring and modulation format identification using asynchronous single channel sampling,” Opt. Express 23(23), 30337–30346 (2015). [CrossRef]   [PubMed]  

12. D. Dahan, D. Levy, and U. Mahlab, “Low cost multi-impairment monitoring technique for 43 Gbps DPSK and 86 Gbps DP-DPSK using delay tap asynchronous sampling method,” in European Conference on Optical Communication (IEEE, 2009), pp. 1–2.

13. R. Borkowski, D. Zibar, A. Caballero, V. Arlunno, and I. T. Monroy, “Stokes space-based optical modulation format recognition for digital coherent receivers,” IEEE Photonic. Tech. L. 25(21), 2129–2132 (2013). [CrossRef]  

14. S. M. Bilal, G. Bosco, Z. Dong, A. P. T. Lau, and C. Lu, “Blind modulation format identification for digital coherent receivers,” Opt. Express 23(20), 26769–26778 (2015). [CrossRef]   [PubMed]  

15. S. Fu, Z. Xu, J. Lu, H. Jiang, Q. Wu, Z. Hu, M. Tang, D. Liu, and C. C.-K. Chan, “Modulation format identification enabled by the digital frequency-offset loading technique for hitless coherent transceiver,” Opt. Express 26(6), 7288–7296 (2018). [CrossRef]   [PubMed]  

16. C. M. Bishop, Pattern recognition and machine learning (Springer, 2006).

17. F. N. Khan, K. Zhong, X. Zhou, W. H. Al-Arashi, C. Yu, C. Lu, and A. P. T. Lau, “Joint OSNR monitoring and modulation format identification in digital coherent receivers using deep neural networks,” Opt. Express 25(15), 17767–17776 (2017). [CrossRef]   [PubMed]  

18. J. Thrane, J. Wass, M. Piels, J. C. Diniz, R. Jones, and D. Zibar, “Machine learning techniques for optical performance monitoring from directly detected PDM-QAM signals,” J. Lightwave Technol. 35(4), 868–875 (2017). [CrossRef]  

19. L. Guesmi, A. M. Ragheb, H. Fathallah, and M. Menif, “Experimental Demonstration of Simultaneous Modulation Format/Symbol Rate Identification and Optical Performance Monitoring for Coherent Optical Systems,” J. Lightwave Technol. 36(11), 2230–2239 (2018). [CrossRef]  

20. D. Wang, M. Zhang, J. Li, Z. Li, J. Li, C. Song, and X. Chen, “Intelligent constellation diagram analyzer using convolutional neural network-based deep learning,” Opt. Express 25(15), 17150–17166 (2017). [CrossRef]   [PubMed]  

21. T. Tanimura, T. Hoshida, T. Kato, S. Watanabe, and H. Morikawa, “Convolutional Neural Network-Based Optical Performance Monitoring for Optical Transport Networks,” J. Opt. Commun. Netw. 11(1), A52–A59 (2019). [CrossRef]  

22. C. Wang, S. Fu, H. Wu, M. Luo, X. Li, M. Tang, and D. Liu, “Joint OSNR and CD monitoring in digital coherent receiver using long short-term memory neural network,” Opt. Express 27(5), 6936–6945 (2019). [CrossRef]   [PubMed]  

23. Z. Wang, A. Yang, P. Guo, and P. He, “OSNR and nonlinear noise power estimation for optical fiber communication systems using LSTM based deep learning technique,” Opt. Express 26(16), 21346–21357 (2018). [CrossRef]   [PubMed]  

24. X. Fan, Y. Xie, F. Ren, Y. Zhang, X. Huang, W. Chen, T. Zhangsun, and J. Wang, “Joint Optical Performance Monitoring and Modulation Format/Bit-Rate Identification by CNN-Based Multi-Task Learning,” IEEE Photonics J. 10(5), 1–12 (2018). [CrossRef]  

25. S. Ruder, “An overview of multi-task learning in deep neural networks,” arXiv preprint arXiv:1706.05098 (2017).

26. F. Agostinelli, M. Hoffman, P. Sadowski, and P. Baldi, “Learning activation functions to improve deep neural networks,” arXiv preprint arXiv:1412.6830 (2014).

27. M. Abadi, P. Barham, J. Chen, Z. Chen, A. Davis, J. Dean, M. Devin, S. Ghemawat, G. Irving, and M. Isard, “Tensorflow: A system for large-scale machine learning,” in 12th {USENIX} Symposium on Operating Systems Design and Implementation ({OSDI} 16)(2016), pp. 265–283.

References

  • View by:
  • |
  • |
  • |

  1. F. N. Khan, Z. Dong, C. Lu, A. P. T. Lau, X. Zhou, and C. Xie, Optical performance monitoring for fiber-optic communication networks (Wiley, 2016).
  2. Z. Dong, F. N. Khan, Q. Sui, K. Zhong, C. Lu, and A. P. T. Lau, “Optical performance monitoring: A review of current and future technologies,” J. Lightwave Technol. 34(2), 525–543 (2016).
    [Crossref]
  3. M. R. Chitgarha, S. Khaleghi, W. Daab, A. Almaiman, M. Ziyadi, A. Mohajerin-Ariaei, D. Rogawski, M. Tur, J. D. Touch, V. Vusirikala, W. Zhao, and A. E. Willner, “Demonstration of in-service wavelength division multiplexing optical-signal-to-noise ratio performance monitoring and operating guidelines for coherent data channels with different modulation formats and various baud rates,” Opt. Lett. 39(6), 1605–1608 (2014).
    [Crossref] [PubMed]
  4. S. Oda, J. Yang, Y. Akasaka, K. Sone, Y. Aoki, M. Sekiya, and J. C. Rasmussen, “In-band OSNR monitor using an optical bandpass filter and optical power measurements for superchannel signals,” in European Conference on Optical Communication (London, U.K), 1–3 (2013).
    [Crossref]
  5. R. Schmogrow, B. Nebendahl, M. Winter, A. Josten, D. Hillerkuss, S. Koenig, J. Meyer, M. Dreschmann, M. Huebner, C. Koos, J. Becker, W. Freude, and J. Leuthold, “Error vector magnitude as a performance measure for advanced modulation formats,” IEEE Photonic. Tech. L. 24(1), 61–63 (2012).
    [Crossref]
  6. C. Do, A. V. Tran, C. Zhu, D. Hewitt, and E. Skafidas, “Data-aided OSNR estimation for QPSK and 16-QAM coherent optical system,” IEEE Photonics J. 5(5), 6601609 (2013).
    [Crossref]
  7. J. H. Lee, D. K. Jung, C. H. Kim, and Y. C. Chung, “OSNR monitoring technique using polarization-nulling method,” IEEE Photonic. Tech. L. 13(1), 88–90 (2001).
    [Crossref]
  8. O. Gerstel, M. Jinno, A. Lord, and S. B. Yoo, “Elastic optical networking: A new dawn for the optical layer?” IEEE Commun. Mag. 50(2), s12–s20 (2012).
    [Crossref]
  9. F. N. Khan, Y. Zhou, A. P. T. Lau, and C. Lu, “Modulation format identification in heterogeneous fiber-optic networks using artificial neural networks,” Opt. Express 20(11), 12422–12431 (2012).
    [Crossref] [PubMed]
  10. Z. Wan, Z. Yu, L. Shu, Y. Zhao, H. Zhang, and K. Xu, “Intelligent optical performance monitor using multi-task learning based artificial neural network,” Opt. Express 27(8), 11281–11291 (2019).
    [Crossref] [PubMed]
  11. F. N. Khan, Y. Yu, M. C. Tan, W. H. Al-Arashi, C. Yu, A. P. T. Lau, and C. Lu, “Experimental demonstration of joint OSNR monitoring and modulation format identification using asynchronous single channel sampling,” Opt. Express 23(23), 30337–30346 (2015).
    [Crossref] [PubMed]
  12. D. Dahan, D. Levy, and U. Mahlab, “Low cost multi-impairment monitoring technique for 43 Gbps DPSK and 86 Gbps DP-DPSK using delay tap asynchronous sampling method,” in European Conference on Optical Communication (IEEE, 2009), pp. 1–2.
  13. R. Borkowski, D. Zibar, A. Caballero, V. Arlunno, and I. T. Monroy, “Stokes space-based optical modulation format recognition for digital coherent receivers,” IEEE Photonic. Tech. L. 25(21), 2129–2132 (2013).
    [Crossref]
  14. S. M. Bilal, G. Bosco, Z. Dong, A. P. T. Lau, and C. Lu, “Blind modulation format identification for digital coherent receivers,” Opt. Express 23(20), 26769–26778 (2015).
    [Crossref] [PubMed]
  15. S. Fu, Z. Xu, J. Lu, H. Jiang, Q. Wu, Z. Hu, M. Tang, D. Liu, and C. C.-K. Chan, “Modulation format identification enabled by the digital frequency-offset loading technique for hitless coherent transceiver,” Opt. Express 26(6), 7288–7296 (2018).
    [Crossref] [PubMed]
  16. C. M. Bishop, Pattern recognition and machine learning (Springer, 2006).
  17. F. N. Khan, K. Zhong, X. Zhou, W. H. Al-Arashi, C. Yu, C. Lu, and A. P. T. Lau, “Joint OSNR monitoring and modulation format identification in digital coherent receivers using deep neural networks,” Opt. Express 25(15), 17767–17776 (2017).
    [Crossref] [PubMed]
  18. J. Thrane, J. Wass, M. Piels, J. C. Diniz, R. Jones, and D. Zibar, “Machine learning techniques for optical performance monitoring from directly detected PDM-QAM signals,” J. Lightwave Technol. 35(4), 868–875 (2017).
    [Crossref]
  19. L. Guesmi, A. M. Ragheb, H. Fathallah, and M. Menif, “Experimental Demonstration of Simultaneous Modulation Format/Symbol Rate Identification and Optical Performance Monitoring for Coherent Optical Systems,” J. Lightwave Technol. 36(11), 2230–2239 (2018).
    [Crossref]
  20. D. Wang, M. Zhang, J. Li, Z. Li, J. Li, C. Song, and X. Chen, “Intelligent constellation diagram analyzer using convolutional neural network-based deep learning,” Opt. Express 25(15), 17150–17166 (2017).
    [Crossref] [PubMed]
  21. T. Tanimura, T. Hoshida, T. Kato, S. Watanabe, and H. Morikawa, “Convolutional Neural Network-Based Optical Performance Monitoring for Optical Transport Networks,” J. Opt. Commun. Netw. 11(1), A52–A59 (2019).
    [Crossref]
  22. C. Wang, S. Fu, H. Wu, M. Luo, X. Li, M. Tang, and D. Liu, “Joint OSNR and CD monitoring in digital coherent receiver using long short-term memory neural network,” Opt. Express 27(5), 6936–6945 (2019).
    [Crossref] [PubMed]
  23. Z. Wang, A. Yang, P. Guo, and P. He, “OSNR and nonlinear noise power estimation for optical fiber communication systems using LSTM based deep learning technique,” Opt. Express 26(16), 21346–21357 (2018).
    [Crossref] [PubMed]
  24. X. Fan, Y. Xie, F. Ren, Y. Zhang, X. Huang, W. Chen, T. Zhangsun, and J. Wang, “Joint Optical Performance Monitoring and Modulation Format/Bit-Rate Identification by CNN-Based Multi-Task Learning,” IEEE Photonics J. 10(5), 1–12 (2018).
    [Crossref]
  25. S. Ruder, “An overview of multi-task learning in deep neural networks,” arXiv preprint arXiv:1706.05098 (2017).
  26. F. Agostinelli, M. Hoffman, P. Sadowski, and P. Baldi, “Learning activation functions to improve deep neural networks,” arXiv preprint arXiv:1412.6830 (2014).
  27. M. Abadi, P. Barham, J. Chen, Z. Chen, A. Davis, J. Dean, M. Devin, S. Ghemawat, G. Irving, and M. Isard, “Tensorflow: A system for large-scale machine learning,” in 12th {USENIX} Symposium on Operating Systems Design and Implementation ({OSDI} 16)(2016), pp. 265–283.

2019 (3)

2018 (4)

2017 (3)

2016 (1)

2015 (2)

2014 (1)

2013 (2)

C. Do, A. V. Tran, C. Zhu, D. Hewitt, and E. Skafidas, “Data-aided OSNR estimation for QPSK and 16-QAM coherent optical system,” IEEE Photonics J. 5(5), 6601609 (2013).
[Crossref]

R. Borkowski, D. Zibar, A. Caballero, V. Arlunno, and I. T. Monroy, “Stokes space-based optical modulation format recognition for digital coherent receivers,” IEEE Photonic. Tech. L. 25(21), 2129–2132 (2013).
[Crossref]

2012 (3)

O. Gerstel, M. Jinno, A. Lord, and S. B. Yoo, “Elastic optical networking: A new dawn for the optical layer?” IEEE Commun. Mag. 50(2), s12–s20 (2012).
[Crossref]

F. N. Khan, Y. Zhou, A. P. T. Lau, and C. Lu, “Modulation format identification in heterogeneous fiber-optic networks using artificial neural networks,” Opt. Express 20(11), 12422–12431 (2012).
[Crossref] [PubMed]

R. Schmogrow, B. Nebendahl, M. Winter, A. Josten, D. Hillerkuss, S. Koenig, J. Meyer, M. Dreschmann, M. Huebner, C. Koos, J. Becker, W. Freude, and J. Leuthold, “Error vector magnitude as a performance measure for advanced modulation formats,” IEEE Photonic. Tech. L. 24(1), 61–63 (2012).
[Crossref]

2001 (1)

J. H. Lee, D. K. Jung, C. H. Kim, and Y. C. Chung, “OSNR monitoring technique using polarization-nulling method,” IEEE Photonic. Tech. L. 13(1), 88–90 (2001).
[Crossref]

Al-Arashi, W. H.

Almaiman, A.

Arlunno, V.

R. Borkowski, D. Zibar, A. Caballero, V. Arlunno, and I. T. Monroy, “Stokes space-based optical modulation format recognition for digital coherent receivers,” IEEE Photonic. Tech. L. 25(21), 2129–2132 (2013).
[Crossref]

Becker, J.

R. Schmogrow, B. Nebendahl, M. Winter, A. Josten, D. Hillerkuss, S. Koenig, J. Meyer, M. Dreschmann, M. Huebner, C. Koos, J. Becker, W. Freude, and J. Leuthold, “Error vector magnitude as a performance measure for advanced modulation formats,” IEEE Photonic. Tech. L. 24(1), 61–63 (2012).
[Crossref]

Bilal, S. M.

Borkowski, R.

R. Borkowski, D. Zibar, A. Caballero, V. Arlunno, and I. T. Monroy, “Stokes space-based optical modulation format recognition for digital coherent receivers,” IEEE Photonic. Tech. L. 25(21), 2129–2132 (2013).
[Crossref]

Bosco, G.

Caballero, A.

R. Borkowski, D. Zibar, A. Caballero, V. Arlunno, and I. T. Monroy, “Stokes space-based optical modulation format recognition for digital coherent receivers,” IEEE Photonic. Tech. L. 25(21), 2129–2132 (2013).
[Crossref]

Chan, C. C.-K.

Chen, W.

X. Fan, Y. Xie, F. Ren, Y. Zhang, X. Huang, W. Chen, T. Zhangsun, and J. Wang, “Joint Optical Performance Monitoring and Modulation Format/Bit-Rate Identification by CNN-Based Multi-Task Learning,” IEEE Photonics J. 10(5), 1–12 (2018).
[Crossref]

Chen, X.

Chitgarha, M. R.

Chung, Y. C.

J. H. Lee, D. K. Jung, C. H. Kim, and Y. C. Chung, “OSNR monitoring technique using polarization-nulling method,” IEEE Photonic. Tech. L. 13(1), 88–90 (2001).
[Crossref]

Daab, W.

Dahan, D.

D. Dahan, D. Levy, and U. Mahlab, “Low cost multi-impairment monitoring technique for 43 Gbps DPSK and 86 Gbps DP-DPSK using delay tap asynchronous sampling method,” in European Conference on Optical Communication (IEEE, 2009), pp. 1–2.

Diniz, J. C.

Do, C.

C. Do, A. V. Tran, C. Zhu, D. Hewitt, and E. Skafidas, “Data-aided OSNR estimation for QPSK and 16-QAM coherent optical system,” IEEE Photonics J. 5(5), 6601609 (2013).
[Crossref]

Dong, Z.

Dreschmann, M.

R. Schmogrow, B. Nebendahl, M. Winter, A. Josten, D. Hillerkuss, S. Koenig, J. Meyer, M. Dreschmann, M. Huebner, C. Koos, J. Becker, W. Freude, and J. Leuthold, “Error vector magnitude as a performance measure for advanced modulation formats,” IEEE Photonic. Tech. L. 24(1), 61–63 (2012).
[Crossref]

Fan, X.

X. Fan, Y. Xie, F. Ren, Y. Zhang, X. Huang, W. Chen, T. Zhangsun, and J. Wang, “Joint Optical Performance Monitoring and Modulation Format/Bit-Rate Identification by CNN-Based Multi-Task Learning,” IEEE Photonics J. 10(5), 1–12 (2018).
[Crossref]

Fathallah, H.

Freude, W.

R. Schmogrow, B. Nebendahl, M. Winter, A. Josten, D. Hillerkuss, S. Koenig, J. Meyer, M. Dreschmann, M. Huebner, C. Koos, J. Becker, W. Freude, and J. Leuthold, “Error vector magnitude as a performance measure for advanced modulation formats,” IEEE Photonic. Tech. L. 24(1), 61–63 (2012).
[Crossref]

Fu, S.

Gerstel, O.

O. Gerstel, M. Jinno, A. Lord, and S. B. Yoo, “Elastic optical networking: A new dawn for the optical layer?” IEEE Commun. Mag. 50(2), s12–s20 (2012).
[Crossref]

Guesmi, L.

Guo, P.

He, P.

Hewitt, D.

C. Do, A. V. Tran, C. Zhu, D. Hewitt, and E. Skafidas, “Data-aided OSNR estimation for QPSK and 16-QAM coherent optical system,” IEEE Photonics J. 5(5), 6601609 (2013).
[Crossref]

Hillerkuss, D.

R. Schmogrow, B. Nebendahl, M. Winter, A. Josten, D. Hillerkuss, S. Koenig, J. Meyer, M. Dreschmann, M. Huebner, C. Koos, J. Becker, W. Freude, and J. Leuthold, “Error vector magnitude as a performance measure for advanced modulation formats,” IEEE Photonic. Tech. L. 24(1), 61–63 (2012).
[Crossref]

Hoshida, T.

Hu, Z.

Huang, X.

X. Fan, Y. Xie, F. Ren, Y. Zhang, X. Huang, W. Chen, T. Zhangsun, and J. Wang, “Joint Optical Performance Monitoring and Modulation Format/Bit-Rate Identification by CNN-Based Multi-Task Learning,” IEEE Photonics J. 10(5), 1–12 (2018).
[Crossref]

Huebner, M.

R. Schmogrow, B. Nebendahl, M. Winter, A. Josten, D. Hillerkuss, S. Koenig, J. Meyer, M. Dreschmann, M. Huebner, C. Koos, J. Becker, W. Freude, and J. Leuthold, “Error vector magnitude as a performance measure for advanced modulation formats,” IEEE Photonic. Tech. L. 24(1), 61–63 (2012).
[Crossref]

Jiang, H.

Jinno, M.

O. Gerstel, M. Jinno, A. Lord, and S. B. Yoo, “Elastic optical networking: A new dawn for the optical layer?” IEEE Commun. Mag. 50(2), s12–s20 (2012).
[Crossref]

Jones, R.

Josten, A.

R. Schmogrow, B. Nebendahl, M. Winter, A. Josten, D. Hillerkuss, S. Koenig, J. Meyer, M. Dreschmann, M. Huebner, C. Koos, J. Becker, W. Freude, and J. Leuthold, “Error vector magnitude as a performance measure for advanced modulation formats,” IEEE Photonic. Tech. L. 24(1), 61–63 (2012).
[Crossref]

Jung, D. K.

J. H. Lee, D. K. Jung, C. H. Kim, and Y. C. Chung, “OSNR monitoring technique using polarization-nulling method,” IEEE Photonic. Tech. L. 13(1), 88–90 (2001).
[Crossref]

Kato, T.

Khaleghi, S.

Khan, F. N.

Kim, C. H.

J. H. Lee, D. K. Jung, C. H. Kim, and Y. C. Chung, “OSNR monitoring technique using polarization-nulling method,” IEEE Photonic. Tech. L. 13(1), 88–90 (2001).
[Crossref]

Koenig, S.

R. Schmogrow, B. Nebendahl, M. Winter, A. Josten, D. Hillerkuss, S. Koenig, J. Meyer, M. Dreschmann, M. Huebner, C. Koos, J. Becker, W. Freude, and J. Leuthold, “Error vector magnitude as a performance measure for advanced modulation formats,” IEEE Photonic. Tech. L. 24(1), 61–63 (2012).
[Crossref]

Koos, C.

R. Schmogrow, B. Nebendahl, M. Winter, A. Josten, D. Hillerkuss, S. Koenig, J. Meyer, M. Dreschmann, M. Huebner, C. Koos, J. Becker, W. Freude, and J. Leuthold, “Error vector magnitude as a performance measure for advanced modulation formats,” IEEE Photonic. Tech. L. 24(1), 61–63 (2012).
[Crossref]

Lau, A. P. T.

Lee, J. H.

J. H. Lee, D. K. Jung, C. H. Kim, and Y. C. Chung, “OSNR monitoring technique using polarization-nulling method,” IEEE Photonic. Tech. L. 13(1), 88–90 (2001).
[Crossref]

Leuthold, J.

R. Schmogrow, B. Nebendahl, M. Winter, A. Josten, D. Hillerkuss, S. Koenig, J. Meyer, M. Dreschmann, M. Huebner, C. Koos, J. Becker, W. Freude, and J. Leuthold, “Error vector magnitude as a performance measure for advanced modulation formats,” IEEE Photonic. Tech. L. 24(1), 61–63 (2012).
[Crossref]

Levy, D.

D. Dahan, D. Levy, and U. Mahlab, “Low cost multi-impairment monitoring technique for 43 Gbps DPSK and 86 Gbps DP-DPSK using delay tap asynchronous sampling method,” in European Conference on Optical Communication (IEEE, 2009), pp. 1–2.

Li, J.

Li, X.

Li, Z.

Liu, D.

Lord, A.

O. Gerstel, M. Jinno, A. Lord, and S. B. Yoo, “Elastic optical networking: A new dawn for the optical layer?” IEEE Commun. Mag. 50(2), s12–s20 (2012).
[Crossref]

Lu, C.

Lu, J.

Luo, M.

Mahlab, U.

D. Dahan, D. Levy, and U. Mahlab, “Low cost multi-impairment monitoring technique for 43 Gbps DPSK and 86 Gbps DP-DPSK using delay tap asynchronous sampling method,” in European Conference on Optical Communication (IEEE, 2009), pp. 1–2.

Menif, M.

Meyer, J.

R. Schmogrow, B. Nebendahl, M. Winter, A. Josten, D. Hillerkuss, S. Koenig, J. Meyer, M. Dreschmann, M. Huebner, C. Koos, J. Becker, W. Freude, and J. Leuthold, “Error vector magnitude as a performance measure for advanced modulation formats,” IEEE Photonic. Tech. L. 24(1), 61–63 (2012).
[Crossref]

Mohajerin-Ariaei, A.

Monroy, I. T.

R. Borkowski, D. Zibar, A. Caballero, V. Arlunno, and I. T. Monroy, “Stokes space-based optical modulation format recognition for digital coherent receivers,” IEEE Photonic. Tech. L. 25(21), 2129–2132 (2013).
[Crossref]

Morikawa, H.

Nebendahl, B.

R. Schmogrow, B. Nebendahl, M. Winter, A. Josten, D. Hillerkuss, S. Koenig, J. Meyer, M. Dreschmann, M. Huebner, C. Koos, J. Becker, W. Freude, and J. Leuthold, “Error vector magnitude as a performance measure for advanced modulation formats,” IEEE Photonic. Tech. L. 24(1), 61–63 (2012).
[Crossref]

Piels, M.

Ragheb, A. M.

Ren, F.

X. Fan, Y. Xie, F. Ren, Y. Zhang, X. Huang, W. Chen, T. Zhangsun, and J. Wang, “Joint Optical Performance Monitoring and Modulation Format/Bit-Rate Identification by CNN-Based Multi-Task Learning,” IEEE Photonics J. 10(5), 1–12 (2018).
[Crossref]

Rogawski, D.

Schmogrow, R.

R. Schmogrow, B. Nebendahl, M. Winter, A. Josten, D. Hillerkuss, S. Koenig, J. Meyer, M. Dreschmann, M. Huebner, C. Koos, J. Becker, W. Freude, and J. Leuthold, “Error vector magnitude as a performance measure for advanced modulation formats,” IEEE Photonic. Tech. L. 24(1), 61–63 (2012).
[Crossref]

Shu, L.

Skafidas, E.

C. Do, A. V. Tran, C. Zhu, D. Hewitt, and E. Skafidas, “Data-aided OSNR estimation for QPSK and 16-QAM coherent optical system,” IEEE Photonics J. 5(5), 6601609 (2013).
[Crossref]

Song, C.

Sui, Q.

Tan, M. C.

Tang, M.

Tanimura, T.

Thrane, J.

Touch, J. D.

Tran, A. V.

C. Do, A. V. Tran, C. Zhu, D. Hewitt, and E. Skafidas, “Data-aided OSNR estimation for QPSK and 16-QAM coherent optical system,” IEEE Photonics J. 5(5), 6601609 (2013).
[Crossref]

Tur, M.

Vusirikala, V.

Wan, Z.

Wang, C.

Wang, D.

Wang, J.

X. Fan, Y. Xie, F. Ren, Y. Zhang, X. Huang, W. Chen, T. Zhangsun, and J. Wang, “Joint Optical Performance Monitoring and Modulation Format/Bit-Rate Identification by CNN-Based Multi-Task Learning,” IEEE Photonics J. 10(5), 1–12 (2018).
[Crossref]

Wang, Z.

Wass, J.

Watanabe, S.

Willner, A. E.

Winter, M.

R. Schmogrow, B. Nebendahl, M. Winter, A. Josten, D. Hillerkuss, S. Koenig, J. Meyer, M. Dreschmann, M. Huebner, C. Koos, J. Becker, W. Freude, and J. Leuthold, “Error vector magnitude as a performance measure for advanced modulation formats,” IEEE Photonic. Tech. L. 24(1), 61–63 (2012).
[Crossref]

Wu, H.

Wu, Q.

Xie, Y.

X. Fan, Y. Xie, F. Ren, Y. Zhang, X. Huang, W. Chen, T. Zhangsun, and J. Wang, “Joint Optical Performance Monitoring and Modulation Format/Bit-Rate Identification by CNN-Based Multi-Task Learning,” IEEE Photonics J. 10(5), 1–12 (2018).
[Crossref]

Xu, K.

Xu, Z.

Yang, A.

Yoo, S. B.

O. Gerstel, M. Jinno, A. Lord, and S. B. Yoo, “Elastic optical networking: A new dawn for the optical layer?” IEEE Commun. Mag. 50(2), s12–s20 (2012).
[Crossref]

Yu, C.

Yu, Y.

Yu, Z.

Zhang, H.

Zhang, M.

Zhang, Y.

X. Fan, Y. Xie, F. Ren, Y. Zhang, X. Huang, W. Chen, T. Zhangsun, and J. Wang, “Joint Optical Performance Monitoring and Modulation Format/Bit-Rate Identification by CNN-Based Multi-Task Learning,” IEEE Photonics J. 10(5), 1–12 (2018).
[Crossref]

Zhangsun, T.

X. Fan, Y. Xie, F. Ren, Y. Zhang, X. Huang, W. Chen, T. Zhangsun, and J. Wang, “Joint Optical Performance Monitoring and Modulation Format/Bit-Rate Identification by CNN-Based Multi-Task Learning,” IEEE Photonics J. 10(5), 1–12 (2018).
[Crossref]

Zhao, W.

Zhao, Y.

Zhong, K.

Zhou, X.

Zhou, Y.

Zhu, C.

C. Do, A. V. Tran, C. Zhu, D. Hewitt, and E. Skafidas, “Data-aided OSNR estimation for QPSK and 16-QAM coherent optical system,” IEEE Photonics J. 5(5), 6601609 (2013).
[Crossref]

Zibar, D.

J. Thrane, J. Wass, M. Piels, J. C. Diniz, R. Jones, and D. Zibar, “Machine learning techniques for optical performance monitoring from directly detected PDM-QAM signals,” J. Lightwave Technol. 35(4), 868–875 (2017).
[Crossref]

R. Borkowski, D. Zibar, A. Caballero, V. Arlunno, and I. T. Monroy, “Stokes space-based optical modulation format recognition for digital coherent receivers,” IEEE Photonic. Tech. L. 25(21), 2129–2132 (2013).
[Crossref]

Ziyadi, M.

IEEE Commun. Mag. (1)

O. Gerstel, M. Jinno, A. Lord, and S. B. Yoo, “Elastic optical networking: A new dawn for the optical layer?” IEEE Commun. Mag. 50(2), s12–s20 (2012).
[Crossref]

IEEE Photonic. Tech. L. (3)

R. Schmogrow, B. Nebendahl, M. Winter, A. Josten, D. Hillerkuss, S. Koenig, J. Meyer, M. Dreschmann, M. Huebner, C. Koos, J. Becker, W. Freude, and J. Leuthold, “Error vector magnitude as a performance measure for advanced modulation formats,” IEEE Photonic. Tech. L. 24(1), 61–63 (2012).
[Crossref]

R. Borkowski, D. Zibar, A. Caballero, V. Arlunno, and I. T. Monroy, “Stokes space-based optical modulation format recognition for digital coherent receivers,” IEEE Photonic. Tech. L. 25(21), 2129–2132 (2013).
[Crossref]

J. H. Lee, D. K. Jung, C. H. Kim, and Y. C. Chung, “OSNR monitoring technique using polarization-nulling method,” IEEE Photonic. Tech. L. 13(1), 88–90 (2001).
[Crossref]

IEEE Photonics J. (2)

C. Do, A. V. Tran, C. Zhu, D. Hewitt, and E. Skafidas, “Data-aided OSNR estimation for QPSK and 16-QAM coherent optical system,” IEEE Photonics J. 5(5), 6601609 (2013).
[Crossref]

X. Fan, Y. Xie, F. Ren, Y. Zhang, X. Huang, W. Chen, T. Zhangsun, and J. Wang, “Joint Optical Performance Monitoring and Modulation Format/Bit-Rate Identification by CNN-Based Multi-Task Learning,” IEEE Photonics J. 10(5), 1–12 (2018).
[Crossref]

J. Lightwave Technol. (3)

J. Opt. Commun. Netw. (1)

Opt. Express (9)

C. Wang, S. Fu, H. Wu, M. Luo, X. Li, M. Tang, and D. Liu, “Joint OSNR and CD monitoring in digital coherent receiver using long short-term memory neural network,” Opt. Express 27(5), 6936–6945 (2019).
[Crossref] [PubMed]

Z. Wang, A. Yang, P. Guo, and P. He, “OSNR and nonlinear noise power estimation for optical fiber communication systems using LSTM based deep learning technique,” Opt. Express 26(16), 21346–21357 (2018).
[Crossref] [PubMed]

F. N. Khan, K. Zhong, X. Zhou, W. H. Al-Arashi, C. Yu, C. Lu, and A. P. T. Lau, “Joint OSNR monitoring and modulation format identification in digital coherent receivers using deep neural networks,” Opt. Express 25(15), 17767–17776 (2017).
[Crossref] [PubMed]

D. Wang, M. Zhang, J. Li, Z. Li, J. Li, C. Song, and X. Chen, “Intelligent constellation diagram analyzer using convolutional neural network-based deep learning,” Opt. Express 25(15), 17150–17166 (2017).
[Crossref] [PubMed]

S. M. Bilal, G. Bosco, Z. Dong, A. P. T. Lau, and C. Lu, “Blind modulation format identification for digital coherent receivers,” Opt. Express 23(20), 26769–26778 (2015).
[Crossref] [PubMed]

S. Fu, Z. Xu, J. Lu, H. Jiang, Q. Wu, Z. Hu, M. Tang, D. Liu, and C. C.-K. Chan, “Modulation format identification enabled by the digital frequency-offset loading technique for hitless coherent transceiver,” Opt. Express 26(6), 7288–7296 (2018).
[Crossref] [PubMed]

F. N. Khan, Y. Zhou, A. P. T. Lau, and C. Lu, “Modulation format identification in heterogeneous fiber-optic networks using artificial neural networks,” Opt. Express 20(11), 12422–12431 (2012).
[Crossref] [PubMed]

Z. Wan, Z. Yu, L. Shu, Y. Zhao, H. Zhang, and K. Xu, “Intelligent optical performance monitor using multi-task learning based artificial neural network,” Opt. Express 27(8), 11281–11291 (2019).
[Crossref] [PubMed]

F. N. Khan, Y. Yu, M. C. Tan, W. H. Al-Arashi, C. Yu, A. P. T. Lau, and C. Lu, “Experimental demonstration of joint OSNR monitoring and modulation format identification using asynchronous single channel sampling,” Opt. Express 23(23), 30337–30346 (2015).
[Crossref] [PubMed]

Opt. Lett. (1)

Other (7)

S. Oda, J. Yang, Y. Akasaka, K. Sone, Y. Aoki, M. Sekiya, and J. C. Rasmussen, “In-band OSNR monitor using an optical bandpass filter and optical power measurements for superchannel signals,” in European Conference on Optical Communication (London, U.K), 1–3 (2013).
[Crossref]

D. Dahan, D. Levy, and U. Mahlab, “Low cost multi-impairment monitoring technique for 43 Gbps DPSK and 86 Gbps DP-DPSK using delay tap asynchronous sampling method,” in European Conference on Optical Communication (IEEE, 2009), pp. 1–2.

C. M. Bishop, Pattern recognition and machine learning (Springer, 2006).

F. N. Khan, Z. Dong, C. Lu, A. P. T. Lau, X. Zhou, and C. Xie, Optical performance monitoring for fiber-optic communication networks (Wiley, 2016).

S. Ruder, “An overview of multi-task learning in deep neural networks,” arXiv preprint arXiv:1706.05098 (2017).

F. Agostinelli, M. Hoffman, P. Sadowski, and P. Baldi, “Learning activation functions to improve deep neural networks,” arXiv preprint arXiv:1412.6830 (2014).

M. Abadi, P. Barham, J. Chen, Z. Chen, A. Davis, J. Dean, M. Devin, S. Ghemawat, G. Irving, and M. Isard, “Tensorflow: A system for large-scale machine learning,” in 12th {USENIX} Symposium on Operating Systems Design and Implementation ({OSDI} 16)(2016), pp. 265–283.

Cited By

OSA participates in Crossref's Cited-By Linking service. Citing articles from OSA journals and other participating publishers are listed here.

Alert me when this article is cited.


Figures (14)

Fig. 1
Fig. 1 AAHs with 80 bins for signals with various OSNRs, modulation formats and baud rates after asynchronous sampling.
Fig. 2
Fig. 2 DNN structure for (a) classification and (b) regression.
Fig. 3
Fig. 3 MT-DNN structure with AAHs bins vector as input and identified OSNRs, symbol rates and modulation formats as output.
Fig. 4
Fig. 4 Simulation setup for joint MFI, BRI and OSNR monitoring with a single MT-DNN and directly detected PDM-QAM signals.
Fig. 5
Fig. 5 Variation of MFI, BRI and OSNR monitoring results with respect to λ3 during the testing process.
Fig. 6
Fig. 6 (a) MFI and BRI accuracy versus epoch during the training process, and (b) OSNR monitoring error for all signals with three modulation formats and two baud rates during the testing process of back-to-back transmission system.
Fig. 7
Fig. 7 (a) RMSE of DNN-based OSNR monitoring with respect to four modulation formats, and (b) OSNR monitoring error for 64QAM signals during the testing process.
Fig. 8
Fig. 8 (a) BRI accuracy for three modulation formats during the testing process, and (b) BRI accuracy versus symbol rate of QPSK signals during the testing process.
Fig. 9
Fig. 9 (a) Variation of BRI and OSNR monitoring results with respect to CD for 20Gbaud and 30Gbaud PDM-16QAM signals during the testing process. (b) AAHs with 80 bins for 30Gbaud PDM-16QAM signals with various CDs (I) and OSNRs (II).
Fig. 10
Fig. 10 (a) Variation of BRI and OSNR monitoring results with respect to DGD for 20Gbaud and 30Gbaud PDM-16QAM signals during the testing process, and (b) AAHs for 30Gbaud PDM-16QAM with various DGDs (0ps,10ps) after asynchronous sampling (OSNR is 26dB).
Fig. 11
Fig. 11 (a) MFI accuracy versus epoch during the training process with a ST-DNN, and (b) MFI and BRI accuracy versus epoch during the training process with a MT-DNN. (c) AAHs for signals at 1600ps/nm CD with various formats after the asynchronous sampling.
Fig. 12
Fig. 12 OSNR monitoring error for all signals at 1600ps/nm CD and 6ps DGD during the testing process.
Fig. 13
Fig. 13 Experimental setup of MFI, BRI and OSNR monitoring with a single MT-DNN and directly detected PDM-QAM signals.
Fig. 14
Fig. 14 (a) BRI accuracy for three modulation formats during the testing process, and (b) OSNR monitoring error of all testing data set in the experiment.

Tables (2)

Tables Icon

Table 1 MFI accuracies for three modulation formats during the testing process under the condition of back-to-back transmission

Tables Icon

Table 2 MFI accuracies for various modulation formats during the testing process in the experiment

Equations (5)

Equations on this page are rendered with MathJax. Learn more.

Softmax( x i )= e x i i m e x i
L= λ 1 L 1 + λ 2 L 2 + λ 3 L 3
L 1 = L 2 = 1 m [ i=1 m y i log y i +(1 y i )log( 1 y i ) ]
L 3 = 1 m i=1 m ( y i y i ) 2
RMSE= i=1 n ( y i y i ) 2 n

Metrics