Development of Audio Source Separation Algorithm in Noisy Environment Using Compact Kernel Time-Frequency Distribution

This research focuses on the development of a source separation algorithm tailored to significantly enhance audio processing in noisy environments. By utilising advanced signal processing techniques and algorithms based on time-frequency analysis, the study explores the effectiveness of the Compact Kernel Distribution (CKD) for this purpose. Performance was evaluated using key metrics such as Signal-to-Interference Ratio (SIR), Source-to- Distortion Ratio (SDR), and Signal-to-Noise Ratio (SNR). Notable improvements were observed: SIR improved by 4.90% and decreased by 5.36%, while SDR improved by 66.47% and 58.08% at SNR of 5 dB SNR for the audio recordings signals compared to Max-Corr and Simplex-Corr, respectively. The results demonstrate that the developed algorithm significantly enhances SIR, SDR, and SNR metrics, with potential applications across various industries, including speech enhancement.