Recognizing Speech Commands Using Convolutional Neural Network

Voice recognition is an important element of modern life and devices used. This is now an integral part of human-computer interfaces. Voice recognition is present in personal assistants like Google Assistant, Microsoft Cortana, Amazon Alexa and Apple Siri to self-driving car and many other industries. In this work a different approach to recognizing voice commands has been shown. We present a new idea to speech recognition, based on the specific coding of time characteristics and frequency characteristics of speech, using convolutional neural networks. Our idea assumes creating patterns for sounds in the form of RGB images and grayscale images. The studies, carried out for isolated words, show high efficacy recognize voice commands.