A Neural State-Space Model Approach to Efficient Speech Separation
This work addresses efficient speech separation for audio processing applications, presenting an incremental improvement in model efficiency.
The authors tackled speech separation by introducing S4M, a neural state-space model framework that efficiently models input signals as linear ODEs for representation learning, achieving comparable SI-SDRi to other backbones with significantly lower complexity, such as a 1.8M-parameter model surpassing a 26.0M-parameter one in noisy conditions.
In this work, we introduce S4M, a new efficient speech separation framework based on neural state-space models (SSM). Motivated by linear time-invariant systems for sequence modeling, our SSM-based approach can efficiently model input signals into a format of linear ordinary differential equations (ODEs) for representation learning. To extend the SSM technique into speech separation tasks, we first decompose the input mixture into multi-scale representations with different resolutions. This mechanism enables S4M to learn globally coherent separation and reconstruction. The experimental results show that S4M performs comparably to other separation backbones in terms of SI-SDRi, while having a much lower model complexity with significantly fewer trainable parameters. In addition, our S4M-tiny model (1.8M parameters) even surpasses attention-based Sepformer (26.0M parameters) in noisy conditions with only 9.2 of multiply-accumulate operation (MACs).