Skip to Main Content
A blind source separation (BSS) algorithm based on time frequency masking is presented here. The algorithm is suitable for sources that are sparse both in the time and frequency domains. The filtering is performed in two steps. In the first step, the signals are subjected to frequency domain filtering where the frequency compositions of the sources are modeled as a mixture of Gaussians. In the next step, the signal is further filtered in the time domain where certain high energy portions of the signals are removed. The success rate of the algorithm depends on the sparsity of the sources in the time and frequency domains. The algorithm has successfully separated foetal and mother's heartbeat sounds from a mixture. Only one sensor signal was used in the separation process.