Work

Multi-Resolution Location-Based Training for Multi-Channel Continuous Speech Separation

Year: 2023

Type: conference-paper

Abstract: The performance of automatic speech recognition (ASR) systems severely degrades when multi-talker speech overlap occurs. In meeting environments, speech separation is typically performed to improve the robustness of ASR systems. Recently, location-based training (LBT) was proposed as a new training criterion for multi-channel talker-independent speaker separation. Assuming fixed array geometry, LBT outperforms widely-used permutation-invariant training in fully overlapped utterances and matched reverberant conditions. This paper extends LBT to conversational multi-channel speaker separation. We introduce multi-resolution LBT to estimate the complex spectrograms from low to high time and frequency resolutions. With multi-resolution LBT, convolutional kernels are assigned consistently based on speaker locations in physical space. Evaluation results show that multi-resolution LBT consistently outperforms other competitive methods on the recorded LibriCSS corpus.

Source: IEEE International Conference on Acoustics Speech and Signal Processing

DOI: 10.1109/icassp49357.2023.10096684

Authors: Hassan Taherian, DeLiang Wang

Institution: The Ohio State University

Language: English

Topic: Speech and Audio Processing

Subfield: Signal Processing

Field: Computer Science

Domain: Physical Sciences

Open Access status: green

Funders: National Science Foundation, Ohio Supercomputer Center, Division of Electrical, Communications and Cyber Systems

Awards: Category I: Bridges-2: Scalable Converged Computing, Data, and Analytics for Rapidly Evolving Science and Engineering Research, Deep Learning Based Complex Spectral Mapping for Multi-Channel Speaker Separation and Speech Enhancement, Deep neural networks for multi-channel speaker localization and speech separation