Discrimination Module for Voice/Audio Signals Based on Wavelet Ridges Analysis

摘要

Low bit-rate at high quality perception is the aim of coding schemes. Traditionally audio and voice coders have evolved as different paradigms: the state of the art voice-coders cores resides in Algebraic Code Excited Linear Prediction (ACELP) technologies while audio coding has its core in Transform Coding (TC). The Unified Speech-Audio Coding (USAC) scheme has become a new paradigm where the principal goal is to choose between the ACELP or TC to reduce the bit rate and increase the high quality perception. This modern coder is based in a module that decides which core coder to use on a specific signal frame. This paper proposes a decision module based on ridges detection in the wavelet transform of the input signal. Wavelet ridges permit to track the instantaneous frequencies contained in the analyzed signal. These instantaneous frequencies, linked to the signal pitch and its harmonics, permit to establish a module for determining whether it is a voice signal or audio.