Conv-TasNet: Surpassing Ideal Time–Frequency Magnitude Masking for Speech Separation

Short summary

Conv-TasNet, a fully-convolutional time-domain deep learning network, achieves state-of-the-art speech separation by operating directly on the waveform, outperforming time-frequency masking methods and even ideal masks in two-speaker scenarios.

AI-generated from the title and abstract; the full text is not read.