Robust Feature Extraction Techniques in Speech Recognition: A Comparative Analysis

Abstract

As the world is moving towards new era known as the era of ‘Artificial intelligence’ where many of things will be controlled automatically through many sources such as face and thumb lock like this we can control things through sound as people are trying to do so and this thing getting hot day by day but it is not explored that much, in this paper we are exploring sound and its feature extraction techniques through which we can extract features from various types of sound and can make them applicable as this paper presents a survey on feature extraction to comparative analysis with respect to properties such as noisy data, complexity, accuracy, extraction method it will be helpful to use which data set with which type of sound. Feature extractions process has a direct relation with any of the machine learning algorithm. If feature extracted is robust, the use of underlining machine learning algorithm will increase accuracy. This paper targeted only the comparative analysis of features used in literature for sound. In future, two or more features will be combined to enhance the impact of sound recognition systems.