A hybrid scheme for encoding audio signal using hidden Markov models of waveforms

TL;DR

This paper presents a hybrid encoding scheme for audio signals that combines time-scale and time-frequency transforms with hidden Markov models to improve the representation of tonal and transient components.

Contribution

It introduces a novel hybrid approach using hidden Markov models for encoding audio signals with structured approximations of tonal and transient parts.

Findings

Effective separation of tonal and transient components

Improved rate estimates for audio encoding

Enhanced encoding accuracy for audiophonic signals

Abstract

This paper reports on recent results related to audiophonic signals encoding using time-scale and time-frequency transform. More precisely, non-linear, structured approximations for tonal and transient components using local cosine and wavelet bases will be described, yielding expansions of audio signals in the form tonal + transient + residual. We describe a general formulation involving hidden Markov models, together with corresponding rate estimates. Estimators for the balance transient/tonal are also discussed.