Limited error based event localizing temporal decomposition and its application to variable-rate speech coding

Abstract

This paper proposes a novel algorithm for temporal decomposition (TD) of speech, called ‘limited error based event localizing temporal decomposition’ (LEBEL-TD), and its application to variable-rate speech coding. In previous work with TD, TD analysis was usually performed on each speech segment of about 200–300 ms or more, making it impractical for online applications. In this present work, the event localization is determined based on a limited error criterion and a local optimization strategy, which results in an average algorithmic delay of 65 ms. Simulation results show that an average log spectral distortion of about 1.5 dB can be achievable at an event rate of 20 events/s. Also, LEBEL-TD uses neither the computationally costly singular value decomposition routine nor the event refinement process, thus reducing significantly the computational cost of TD. Further, a method for variable-rate speech coding an average rate of around 1.8 kbps based on STRAIGHT (Speech Transformation and Representation using Adaptive Interpolation of weiGHTed spectrum), which is a high-quality speech analysis–synthesis framework, using LEBEL-TD is also realized. Subjective test results indicate that the performance of the proposed speech coding method is comparable to that of the 4.8 kbps FS-1016 CELP coder.