Deep learning-based accompaniment extraction method and system, storage medium and equipment

The invention discloses an accompaniment extraction method and system based on deep learning, a storage medium and equipment, and belongs to the technical field of short-distance wireless communication, and the method comprises the steps: carrying out the framing of a song PCM signal at a wireless transmitting end, and carrying out the windowing according to the frames, and obtaining a windowing signal; performing time-frequency transformation on the windowed signal by using improved discrete cosine transform to obtain an MDCT spectral coefficient; performing feature extraction on the windowed signal to obtain an MDFT amplitude spectrum corresponding to the windowed signal; inputting the MDFT amplitude spectrum into a pre-trained neural network model to obtain a floating value mask; performing point multiplication on the MDCT spectral coefficient and the floating value mask to obtain the spectral coefficient of the accompaniment signal; according to the spectrum coefficient of the accompaniment signal, contin