G.723.1 ALGORITHM USING EFFICIENT VOICE ACTION DETECTOR

摘要

PURPOSE: A G.723.1 algorithm using an efficient voice action detector is provided to judge whether a voice with respect to each frame of 30 ms is present. CONSTITUTION: An energy and a pitch gain are used to discriminate a vocal sound. An energy value obtained from 5 previously vocal sound frames is used as an energy threshold value. When an energy value of a current frame is greater than the energy threshold value, it is discriminated as a vocal sound section. A ZCR parameter of a voiceless is greater than those and a vocal sound and a mute. When a ZCR of a current frame becomes greater than a ZCR threshold value, it is discriminated as a voiceless sound. LSP coefficients has different distribution features in a vocal sound section and a mute section from each other.