Enhancement of speech through source separation for conferencing systems

Speech improvement is an indispensable technology in the field of speech interaction. Speech enhancement uses a variety of techniques and algorithms to increase the quality and intelligibility of speech. Background noise is always present when there is a speech signal. To retrieve the intended speech signal from the damaged speech signal, speech systems must use excellent noise reduction algorithms. This article presents an application-oriented approach for separating and enhancing preferred speech signals using circular microphone array for audio conferencing systems. The source separation for conferencing (SSC) method calculates the number of active speech signals and the direction of arrival by processing the arrival angle, phase variations, and the time difference in arrival. The direction of the signal is determined using Time Difference of Arrival (TDOA) technique. The adaptive least mean square (LMS) algorithm and the transformed TDOA signal can be enhanced to obtain the preferred signal from the selected location. The SSC method yields an SNR of 7.8 dB, SDR of 6.9 dB, SIR of 8.3 dB, and PESQ of 2.2 as the signal quality metrics of an input signal to the enhanced desired signal. The efficiency of the proposed model is compared with the existing models like PL-CNN, DL-MVDR, and SUDoRMRF.