RECOGNITION SYSTEM FOR SOUND SIGNALS OF NON-SPEECH ORIGIN

Abstract

This work is devoted to the development of software that would allow the recognition of sound signals of non-speech origin. The article reviews the current sound recognition systems, identifies their advantages and disadvantages. The list of the most widespread algorithms which can be used for process of training of model and classification of sounds is resulted. The Gaussian mixed distribution model, which is used to describe the program's sound models, is considered in more detail. An overview of the process of recognition of sound signals and their further classification is given. This program can be used for audio detection of signals that are available in the database. The product can be used on its own or included in other more advanced recognition systems. Using the tools of the C ++ programming language allowed to make it efficient and maintain speed. Statistics of recognition results are also given and a conclusion on the efficiency of the system is made.