Main Article Content
We present a model-based approach to voice activity detection (VAD) for a variety of applications including medical, education, and online video conferencing. The artificial neural network is optimally trained to provide a consistent model by using a Mel Frequency Cepstrum Coefficient) factor derived from clean or noisy speech samples. One of the strategies of neuron networks for artificial intelligence applications where this method can differentiate between abnormal users' sound signals and neural networks. The device first trainers fixed weights on these audios and then gives the output for each format and high speed. The proposed neural network analysis is focused on speech recognition solutions, signal detection by means of angular modulation and modulated techniques detection. The results show that: (i) the proposed artificial neural system classification system delivers fairly high scores under a variety of noise conditions; (ii) in terms of different classification steps, the invented model exceeds other VAD methods.