Long Distance Voice Channel Diagnosis Using Deep Neural Networks

Zhen Qin, Tom Ko, Guangjian Tian


In long distance telephone network, it is time-consuming to detect and locate the problematic devices. Although hints could be given from the types of distortion in the test calls, it is tedious to manually classify the distortion types from a large number of calls. In this paper, we present our work on using a deep neural network-based classifier, to automatically detect and identify the type of distortion which often occurs in long distance calls. We verified our approach with data from real telecommunication networks and the results showed that our approach can achieve an average recall rate of 71% in classification. We believe our method can lead to a huge reduction of manpower and time in long distance voice channel troubleshooting.


 DOI: 10.21437/Interspeech.2018-1428

Cite as: Qin, Z., Ko, T., Tian, G. (2018) Long Distance Voice Channel Diagnosis Using Deep Neural Networks. Proc. Interspeech 2018, 2968-2971, DOI: 10.21437/Interspeech.2018-1428.


@inproceedings{Qin2018,
  author={Zhen Qin and Tom Ko and Guangjian Tian},
  title={Long Distance Voice Channel Diagnosis Using Deep Neural Networks},
  year=2018,
  booktitle={Proc. Interspeech 2018},
  pages={2968--2971},
  doi={10.21437/Interspeech.2018-1428},
  url={http://dx.doi.org/10.21437/Interspeech.2018-1428}
}