Evaluation of VoIP Speech Recognition with Packet Loss Concealment
Article Sidebar
Main Article Content
Abstract
To increase the robustness of Automatic Speech Recognition (ASR) on the IP network (VoIP), we propose in this article the use of packet loss concealment (PLC: Packet Loss Concealment). This method consists of generating a synthetic voice signal intended to replace the missing data, ensuring a smooth transition between the real signal and the synthetic signal. Thus, in this work we have adapted the ITU-I G711 Appendix I recommendation to the G729 codec. For the speech recognition part, we implemented the open source HTK (Hidden Markov Models ToolKit) system, while the packet loss is simulated by a two-state Markov model. The experimental results with speech transcoded with the G729 codec used in VoIP networks show a significant improvement in the recognition rate with the packet loss masking method developed, thus supporting the approach followed in this work.
Article Details
References
Ollason, D., Valtchev, V. and
Woodland, P., “The HTK Book
Version 3.3”, Speech group, Engi- neering Department, Cambridge Uni- versity. April 2005 [2] ITU-T — Recommandation G.729, “Codage de la parole à Skbits/s par prédiction linéaire à excitation par séquences codées à structure algébrique conjuguée (CS-ACELP)”, 1996.
[3] Yong, H. and Jiang, Z, “Implementation of ITU-T G729 Speech
Codec in IP Telephony Gateway”,
Wuhan University Joumal of
Natural Sciences, vol. 5, pp.159-163,
2000.
Milner, B. and Semnani, S. “Robust
Speech Recognition Over Networks”, IEEE International Conference Acous- tics, Speech, and Signal Processing, Vol.3, pp. 1791 — 1794, 2000. Recommandation | UIT-T — G711, “Algorithme Simple de haut qualité pour le masquage des pertes en codage G.711”, Septembre 1999.
Wiley, J., “VoIP voice and fax signal processing”, Published simultaneously in Canada, p.592, 2008
Sommen, P.CW. and Jayasinghe, JAKS. “On Frequency Domain Adaptive Filters using the Overlap-add Method”, IEEE Philips Research Laboratories, pp.28-30, 1988
Nakamura, K., “An Improvement of G.711 PLC Using Sinusoidal _ model”, Proceedings of the IEEE The Inter- national Conference on Computer as a Toll, pp.1670-1673, 2005