A Kalman filtering approach to GMM predictive coding of LSFS for packet loss conditions

Shaminda Subasingha, Manohar Murthi, Søren Vang Andersen

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

Gaussian Mixture Model (GMM)-based vector quantization of Line Spectral Frequencies (LSFs) has gained wide acceptance in speech coding. In predictive coding of LSFs, the GMM approach utilizing Kalman filtering principles to account for quantization noise has been shown to perform better than a baseline GMM Recursive Coder approaches for both clean and packet loss conditions at roughly the same complexity However, the GMM Kalman based predictive coder was not specifically designed for operation in packet loss conditions. In this paper, we show how an initial GMM Kalman predictive coder can be utilized to obtain a robust GMM predictive coder specifically designed to operate in packet loss.In particular, we demonstrate how one can define sets of encoding and decoding modes, and design special Kalman encoding and decoding gains for each set. With this framework,GMM predictive coding design can be viewed as determining the special Kalman gains that minimize the expected minimum mean squared error at the decoder in packet loss conditions.The simulation results demonstrate that the proposed robust Kalman predictive coder achieves better performance than the baseline GMM predictive coders.

Original languageEnglish
Title of host publicationDSP 2009: 16th International Conference on Digital Signal Processing, Proceedings
DOIs
StatePublished - Nov 20 2009
EventDSP 2009:16th International Conference on Digital Signal Processing - Santorini, Greece
Duration: Jul 5 2009Jul 7 2009

Other

OtherDSP 2009:16th International Conference on Digital Signal Processing
CountryGreece
CitySantorini
Period7/5/097/7/09

Fingerprint

Packet loss
Decoding
Speech coding
Vector quantization

Keywords

  • GMM
  • Kalman filtering
  • Speech coding
  • Vector quantization

ASJC Scopus subject areas

  • Computer Vision and Pattern Recognition
  • Signal Processing

Cite this

Subasingha, S., Murthi, M., & Vang Andersen, S. (2009). A Kalman filtering approach to GMM predictive coding of LSFS for packet loss conditions. In DSP 2009: 16th International Conference on Digital Signal Processing, Proceedings [5201111] https://doi.org/10.1109/ICDSP.2009.5201111

A Kalman filtering approach to GMM predictive coding of LSFS for packet loss conditions. / Subasingha, Shaminda; Murthi, Manohar; Vang Andersen, Søren.

DSP 2009: 16th International Conference on Digital Signal Processing, Proceedings. 2009. 5201111.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Subasingha, S, Murthi, M & Vang Andersen, S 2009, A Kalman filtering approach to GMM predictive coding of LSFS for packet loss conditions. in DSP 2009: 16th International Conference on Digital Signal Processing, Proceedings., 5201111, DSP 2009:16th International Conference on Digital Signal Processing, Santorini, Greece, 7/5/09. https://doi.org/10.1109/ICDSP.2009.5201111
Subasingha S, Murthi M, Vang Andersen S. A Kalman filtering approach to GMM predictive coding of LSFS for packet loss conditions. In DSP 2009: 16th International Conference on Digital Signal Processing, Proceedings. 2009. 5201111 https://doi.org/10.1109/ICDSP.2009.5201111
Subasingha, Shaminda ; Murthi, Manohar ; Vang Andersen, Søren. / A Kalman filtering approach to GMM predictive coding of LSFS for packet loss conditions. DSP 2009: 16th International Conference on Digital Signal Processing, Proceedings. 2009.
@inproceedings{375e6a7ed6104281b762be4edc44a389,
title = "A Kalman filtering approach to GMM predictive coding of LSFS for packet loss conditions",
abstract = "Gaussian Mixture Model (GMM)-based vector quantization of Line Spectral Frequencies (LSFs) has gained wide acceptance in speech coding. In predictive coding of LSFs, the GMM approach utilizing Kalman filtering principles to account for quantization noise has been shown to perform better than a baseline GMM Recursive Coder approaches for both clean and packet loss conditions at roughly the same complexity However, the GMM Kalman based predictive coder was not specifically designed for operation in packet loss conditions. In this paper, we show how an initial GMM Kalman predictive coder can be utilized to obtain a robust GMM predictive coder specifically designed to operate in packet loss.In particular, we demonstrate how one can define sets of encoding and decoding modes, and design special Kalman encoding and decoding gains for each set. With this framework,GMM predictive coding design can be viewed as determining the special Kalman gains that minimize the expected minimum mean squared error at the decoder in packet loss conditions.The simulation results demonstrate that the proposed robust Kalman predictive coder achieves better performance than the baseline GMM predictive coders.",
keywords = "GMM, Kalman filtering, Speech coding, Vector quantization",
author = "Shaminda Subasingha and Manohar Murthi and {Vang Andersen}, S{\o}ren",
year = "2009",
month = "11",
day = "20",
doi = "10.1109/ICDSP.2009.5201111",
language = "English",
isbn = "9781424432981",
booktitle = "DSP 2009: 16th International Conference on Digital Signal Processing, Proceedings",

}

TY - GEN

T1 - A Kalman filtering approach to GMM predictive coding of LSFS for packet loss conditions

AU - Subasingha, Shaminda

AU - Murthi, Manohar

AU - Vang Andersen, Søren

PY - 2009/11/20

Y1 - 2009/11/20

N2 - Gaussian Mixture Model (GMM)-based vector quantization of Line Spectral Frequencies (LSFs) has gained wide acceptance in speech coding. In predictive coding of LSFs, the GMM approach utilizing Kalman filtering principles to account for quantization noise has been shown to perform better than a baseline GMM Recursive Coder approaches for both clean and packet loss conditions at roughly the same complexity However, the GMM Kalman based predictive coder was not specifically designed for operation in packet loss conditions. In this paper, we show how an initial GMM Kalman predictive coder can be utilized to obtain a robust GMM predictive coder specifically designed to operate in packet loss.In particular, we demonstrate how one can define sets of encoding and decoding modes, and design special Kalman encoding and decoding gains for each set. With this framework,GMM predictive coding design can be viewed as determining the special Kalman gains that minimize the expected minimum mean squared error at the decoder in packet loss conditions.The simulation results demonstrate that the proposed robust Kalman predictive coder achieves better performance than the baseline GMM predictive coders.

AB - Gaussian Mixture Model (GMM)-based vector quantization of Line Spectral Frequencies (LSFs) has gained wide acceptance in speech coding. In predictive coding of LSFs, the GMM approach utilizing Kalman filtering principles to account for quantization noise has been shown to perform better than a baseline GMM Recursive Coder approaches for both clean and packet loss conditions at roughly the same complexity However, the GMM Kalman based predictive coder was not specifically designed for operation in packet loss conditions. In this paper, we show how an initial GMM Kalman predictive coder can be utilized to obtain a robust GMM predictive coder specifically designed to operate in packet loss.In particular, we demonstrate how one can define sets of encoding and decoding modes, and design special Kalman encoding and decoding gains for each set. With this framework,GMM predictive coding design can be viewed as determining the special Kalman gains that minimize the expected minimum mean squared error at the decoder in packet loss conditions.The simulation results demonstrate that the proposed robust Kalman predictive coder achieves better performance than the baseline GMM predictive coders.

KW - GMM

KW - Kalman filtering

KW - Speech coding

KW - Vector quantization

UR - http://www.scopus.com/inward/record.url?scp=70449586857&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=70449586857&partnerID=8YFLogxK

U2 - 10.1109/ICDSP.2009.5201111

DO - 10.1109/ICDSP.2009.5201111

M3 - Conference contribution

AN - SCOPUS:70449586857

SN - 9781424432981

BT - DSP 2009: 16th International Conference on Digital Signal Processing, Proceedings

ER -