SNR Improvement in Voice Activity Detection

Abstract

Over the years, the aim of error-free continuous speech recognition has remained controversial.Environmental robustness has acquired widespread acceptance as one of the key topics of research in the voice recognition field in recent years. Several methods have been investigated. Window framing, speech feature extractions, feature filtering techniques, and other algorithms have all been examined and shown to improve recognition accuracy. One of the neuron connectivity strategies for artificial intelligence applications is to distinguish between sound signals in form of voice and unvoiced. The device first applies fixed weights to these audios before providing output for each format and high speed. Communication through Voice over Internet Protocol (VoIP) in modern-age speech is attracting many researchers. Voice Activity Detector (VAD) is a technique of distinguishing the articulated portion of speech and the silence component in the original voice signal. In many applications such as speech recognition, voice compression systems, mobile communication, this approach plays an important role for obtaining better speech quality and reducing bandwidth complexity.