Efficient Non-linear Changed Mel-filter Bank VAD Algorithm

This paper introduces efficient non-linear changed mel-filter bank (MFB) voice activity detection (VAD) algorithm. Non-linear changed mel-filter bank outputs improve detection of parts in the speech signal, where vowels, diphthongs and semivowels are present. To make voice activity detection of consonants in the speech signal as good as possible, the hangover and hangbefore criteria are used. For this reason the phoneme duration analysis was made. The duration of vowels, diphthongs and semivowels defines how many frames must be detected as speech, so that it can be decided if hangover and hangbefore criteria will be used at all. The duration of consonants defines how many frames will be used for hangover and hangbefore criteria. Comparative tests were made between the MFB VAD algorithm, where non-linear function was used and where it was not used. The experiments were also made on four VAD algorithms used in the ITU G.729, ITU G.723.1, DSR ETSI ES 202 050, and DSR ETSI ES 202 211 standards. The introduction of non-linear function in to the MFB VAD algorithm reduces errors obtained by incorrect voice activity detection. Key-Words: voice activity detection, signal processing, hangover criterion, hangbefore criterion