Robust Speech Coding

Speech transmission over digital communication links requires source and channel coding schemes which are adapted to the transmission conditions. While in most current systems speech is transmitted over channels with fixed bit rates emerging systems rely on packetized transmission schemes. Examples are Voice Over-IP and Push-to-Talk systems. Our interest is focussed around the estimation of speech coder parameters in the presence of acoustic noise and optimal estimation of disturbed or missing parameters in packetized networks.

The Figures below depict the magnitude of the correlation coefficient of speech spectral parameters (line spectral frequencies or LSF) within one frame of speech (intra-frame correlation) and across two consecutive frames (inter-frame correlation). When (in a packetized network) some of the spectral parameters are received and others are lost the correlation can be used to restore the lost parameters. Such a scheme is shown below.

line spectral frequencies

line spectral frequencies


Martin, R., Malah, D., Cox, R.V., Accardi, A.J. (2004). "A Noise Reduction Preprocessor for Mobile Voice Communication, JASP No.8, pp. 1046-1058."

Martin, R.; Hoelper, C.; Wittke, I.: " Estimation of Missing LSF Parameters Using Gaussian Mixture Models ", IEEE International Conference on Acoustics, Speech and Signal Processing, Salt Lake City, May 2001

Martin, R.; Wittke, I.; Jax, P.: " Optimized Estimation of Spectral Parameters for the Coding of Noisy Speech", IEEE International Conference on Acoustics, Speech and Signal Processing, Istanbul, Turkey, June 2000, Vol. III, pp. 1479-1482

Martin, R.; Kang, H.G.; Cox, R.V.: " Low Delay Analysis/Synthesis Schemes for Joint Speech Enhancement and Low Bit Rate Speech Coding", EUROSPEECH-99, Budapest, Hungary, September 1999, S. 1463-1466

Martin, R.; Cox, R.V.: " New Speech Enhancement and Coding Techniques for Low Bit Rate Speech Coding", IEEE Workshop on Speech Coding, Haikko Manor, Porvoo, Finland, 21.-23. Juni 1999, S. 165-167