Features for voice activity detection: a comparative analysis

Abstract

In many speech signal processing applications, voice activity detection (VAD) plays an essential role for separating an audio stream into time intervals that contain speech activity and time intervals where speech is absent. Many features that reflect the presence of speech were introduced in literature. However, to our knowledge, no extensive comparison has been provided yet. In this article, we therefore present a structured overview of several established VAD features that target at different properties of speech. We categorize the features with respect to properties that are exploited, such as power, harmonicity, or modulation, and evaluate the performance of some dedicated features. The importance of temporal context is discussed in relation to latency restrictions imposed by different applications. Our analyses allow for selecting promising VAD features and finding a reasonable trade-off between performance and complexity.

Efficient target activity detection based on recurrent neural networks

“…A batch of training data comprises 128 sequences, with each sequence consisting of 20 feature vectors. The feature vector x t is computed from the observed signals every millisecond according to (6). The dataset consists of recordings of a desired target speaker, up to 4 simultaneously active interferers, and babble noise in the background.…”

“…In order to achieve a comparable performance both on small network sizes and small amounts of training data, the selection of feature vectors is indispensable. Classical approaches for Voice Activity Detection (VAD) are typically single-channel methods exploiting distinctive properties of speech signals like stationarity, harmonic structure and spectral envelopes in order to differentiate between speech and background noise [5,6]. These VAD methods, however, cannot be used to differentiate between a target speaker and interfering speech sources as the proposed TAD does.…”

Finding Time Together: Detection and Classification of Focused Interaction in Egocentric Video

“…Voice Activity Detection (VAD) is widely researched in audio signal processing and used for audio conferencing, speech encoding, speech recognition, and speaker recognition [17,26]. VAD methods detect voice activity (primarily speech) from a noisy audio signal [16,24,29]. Video content-based camera motion analysis methods make use of template matching [1] and optical flow [6].…”

A priori SNR estimation and noise estimation for speech enhancement

“…It is a fact that SNR can be high at a single frequency point when speech (especially voiced frame) is present, even though the overall SNR of a signal is low (such as 0 dB) [13]. To each frequency point, the entropy of the R continuous frames before the current frame will abruptly become small when speech suddenly appears in the current frame.…”