Abstract:
Overlapped Speech refers to speakers speaking simultaneously (i.e. speech mixture). Speech separation has long been an active research topic in the signal processing community, with its importance in a wide range of applications such as hearable devices and telecommunication systems. It is a fundamental problem for all higher-level speech processing tasks.
With recent progress in deep neural networks, the separation performance has been significantly advanced by various new problems. The problem formulation of time-domain, end-to-end speech separation naturally arises to tackle the disadvantages in frequency-domain systems. We’ve used a dual path recurrent neural network for separation of mixed audios.
DPRNN (Dual-Path RNN) primarily separates in the time domain for audio source separation. It leverages recurrent neural networks (RNNs) to process temporal sequences of audio data. DPRNN focuses on exploiting temporal dependencies within audio signals for effective separation. We looked into the training objectives for separating and improving the robustness under reverberant environments. This project is further analyzed and can be used as the basis for future works.
