Research on sub band filtering approach for multi band spectral subtraction based acoustic noise suppression system

Abstract

In voice communication speech enhancement algorithms are very much essential for effective information exchange. With the development of mobile communication there is a great demand for speech processing devices. The latest research in speech processing techniques particularly speech enhancement is largely driven by real world applications. The development of Digital signal processing system has brought incredible flexibility in speech processing systems by which quality of service is enhanced with less power consumption and reduced cost. newlineVery large scale integration Technology combined with mobile communication, process the speech signals to maintain good quality of speech at both the ends of communication system. Most of the cell phones are smart devices which make use of pre-processing algorithms for speech enhancement. The ability of these smart devices depends on these algorithms. These algorithms must be adaptive in nature to suppress real time noise since cell phones are used under various background conditions. newlineSpeech signals are complex in nature speech hence multiple methods are applied to extract newlinethe information out of it. Generally speech signals are separated into short frames for better estimation of the tonal components. In this thesis we propose new algorithms to assess the signals in small bands based on type of background disturbance as a result filters are adapted to time-varying signals. In our research work, new algorithms are developed and performances of these algorithms are assessed using standard objective parameters SNR, PESQ, SSNR, WSS and LLR. The speech samples corrupted by several real time background noises are derived by Noizeus data base. The data base consists of contains speech uttered by both female and male speakers hence performance of algorithms are compared with both the types of speakers. The use of database helps us to compare the proposed method with the other approved methods for speech enhancement.