Design of a Real-Time Automatic Source Monitoring Framework Based on Sound Source Localization

Abstract

Sound Source Localization is the technique to exactly locate the position of a sound emitting source with the help of the audio data only. Its applications are audio surveillance, security monitoring where visual data is not available, robotics such as robots acting as a waiter, etc. It is also an initial stage of signal processing for the fields of speech enhancement, speech separation and Automatic Speech Recognition (ASR). Localization is usually carried out using array of microphones. The scenario of locating a sound source can be different such as single source localization, multiple source localization, moving source localization, source localization in noisy environment, etc. Extensive research progress is there in this field due to its multidimensional applications. This paper attempts to address some of the existing challenges in this field. Work in this paper deals with designing of a real-time, automatic target monitoring framework based on sound source localization. A comparative study of our results with other standard localization algorithm is also presented. The studies presented in this paper leads to the fact that the designed sound source localization framework is indeed low cost, small, real-time and able to locate multiple sources satisfactorily even in office environment where level of noise and reverberation is higher than standard anechoic rooms.