X-SepFormer: End-to-end Speaker Extraction Network with Explicit Optimization on Speaker Confusion
This work addresses a critical practical issue in speech processing for applications like communication systems, though it is incremental as it builds on existing SepFormer architecture.
The paper tackles the problem of speaker confusion in target speech extraction by proposing two novel loss schemes that optimize for reconstruction improvement at chunk-level, resulting in a 14.8% reduction in speaker confusion errors and achieving state-of-the-art performance with 19.4 dB SI-SDRi and 3.81 PESQ on the WSJ0-2mix dataset.
Target speech extraction (TSE) systems are designed to extract target speech from a multi-talker mixture. The popular training objective for most prior TSE networks is to enhance reconstruction performance of extracted speech waveform. However, it has been reported that a TSE system delivers high reconstruction performance may still suffer low-quality experience problems in practice. One such experience problem is wrong speaker extraction (called speaker confusion, SC), which leads to strong negative experience and hampers effective conversations. To mitigate the imperative SC issue, we reformulate the training objective and propose two novel loss schemes that explore the metric of reconstruction improvement performance defined at small chunk-level and leverage the metric associated distribution information. Both loss schemes aim to encourage a TSE network to pay attention to those SC chunks based on the said distribution information. On this basis, we present X-SepFormer, an end-to-end TSE model with proposed loss schemes and a backbone of SepFormer. Experimental results on the benchmark WSJ0-2mix dataset validate the effectiveness of our proposals, showing consistent improvements on SC errors (by 14.8% relative). Moreover, with SI-SDRi of 19.4 dB and PESQ of 3.81, our best system significantly outperforms the current SOTA systems and offers the top TSE results reported till date on the WSJ0-2mix.