With the improvement of computer performance, audio processing technology has also made tremendous progress. In recent years, edge AI technology has been used in audio signal separation research, becoming an increasingly popular topic in the field of audio signal processing and driving the development of source separation based on deep learning technology. After clarifying the basic theories of music source separation and the preprocessing workflow of audio signals, the study incorporates an attention mechanism and employs a dual-gate mechanism to better control the flow of feature information across different convolutional layers, filtering out unnecessary feature information to achieve effective audio source separation in live music performances. The research results indicate that the proposed algorithm achieves a performance improvement of approximately 4 dB to 10 dB compared to HPSS in terms of SIR values, and at least a 1 dB improvement compared to the REPET algorithm, thereby demonstrating that the proposed method is a more effective separation approach.
