A novel voice activity detection based on phoneme recognition using statistical model

Abstract

In this article, a novel voice activity detection (VAD) approach based on phoneme recognition using Gaussian Mixture Model based Hidden Markov Model (HMM/GMM) is proposed. Some sophisticated speech features such as high order statistics (HOS), harmonic structure information and Mel-frequency cepstral coefficients (MFCCs) are employed to represent each speech/non-speech segment. The main idea of this new method is regarding the non-speech as a new phoneme corresponding to the conventional phonemes in mandarin, and all of them are then trained under maximum likelihood principle with Baum-Welch algorithm using GMM/HMM model. The Viterbi decoding algorithm is finally used for searching the maximum likelihood of the observed signals. The proposed method shows a higher speech/non-speech detection accuracy over a wide range of SNR regimes compared with some existing VAD methods. We also propose a different method to demonstrate that the conventional speech enhancement method only with accurate VAD is not effective enough for automatic speech recognition (ASR) at low SNR regimes.

A Hierarchical Framework Approach for Voice Activity Detection and Speech Enhancement

“…Recently, many advanced techniques were proposed to eliminate the limitation of MFCC. The harmonic structure-related information is robust to high-pitched sounds and seems to be a promising cue to increase robustness of the noise, especially in low SNR conditions [ 13 ]. Fukuda et al [ 8 ] replaced the traditional Mel-frequency cepstral coefficients by the harmonic structure information and made a significant improvement of recognition rate.…”

“…Wu and Zhang [ 12 ] proposed a multiple kernel support vector machine (MK-SVM) method for multiple feature based VAD. Bao and Zhu [ 13 ] combined harmonic structure information and the high order statistics (HOS) with Gaussian mixture model based hidden Markov model (HMM/GMM) for efficient speech/nonspeech classification.…”

Diagnosis method based on hidden Markov model and Weibull mixture model for mechanical faults of in-wheel motors

“…Therefore, choosing a suitable data fitting model is of great significance to the subsequent diagnosis results. The common method of combining data fitting with HMM is Gaussian mixture model (GMM) [17], which is commonly used in the field of speech recognition [22][23][24][25]. The observation probability matrix in the HMM parameters can be described by the GMM of the observation sequence, and the corresponding recognition model can be obtained.…”

Mel-Scaled Autoregressive (Mel-AR) Model based Voice Activity Detection using Likelihood Ratio Measure

“…However, the algorithms based on speech features with heuristic rules have difficulty in coping with real world noises at low SNR conditions. Recently, statistical model based VAD is found to be an efficient approach to segregate speech and non-speech frames under a broad range of background noises [11], [12], [13], [14], [15], [16]. In [11], a robust VAD algorithm based on statistical likelihood ratio test (LRT) involving a single observation vector is proposed.…”