Abstract: Speech is a natural and easiest way of communication. Various researches have been done on voice activity detection and automatic speech recognition system. This paper presents a novel and efficient approach of human voice activity detection in noisy environment. Voice Activity Detection (VAD) is a primary method of speech processing. It refers to a class of methods which detects whether a sound signal contains speech or not i.e. it discriminates human speech signal from other sounds. In real life applications it is not feasible to focus on characteristics of noise due to their random occurrence. Hence to achieve the purpose formants are used as a main features which represents human voice characteristics very effectively. Here clean and noisy speech samples are taken and it is represented that formants of clean speech and noisy speech occurs in same range of frequencies. Thus formants are prominent feature for VAD. To check the robustness of the implemented system clean speech samples are compared with noisy speech samples at various SNR levels.

Keywords: Voice activity detection, Automatic speaker recognition, Formants

| DOI: 10.17148/IJIREEICE.2018.6611