VarArray: Array-Geometry-Agnostic Continuous Speech Separation
Authors:
- 23 May 2022
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 6027-6031
Abstract
Continuous speech separation using a microphone array was shown to be promising in dealing with the speech overlap problem in natural conversation transcription. This paper proposes VarArray, an array-geometry-agnostic speech separation neural network model. The proposed model is applicable to any number of microphones without retraining while leveraging the nonlinear correlation between the input channels. The proposed method adapts different elements that were proposed before separately, including transform-average-concatenate, conformer speech separation, and inter-channel phase differences, and combines them in an efficient and cohesive way. Large-scale evaluation was performed with two real meeting transcription tasks by using a fully developed transcription system requiring no prior knowledge such as reference segmentations, which allowed us to measure the impact that the continuous speech separation system could have in realistic settings. The proposed model outperformed a previous approach to array-geometry-agnostic modeling for all of the geometry configurations considered, achieving asclite-based speaker-agnostic word error rates of 17.5% and 20.4% for the AMI development and evaluation sets, respectively, in the end-to-end setting using no ground-truth segmentations.Keywords
All Related Versions
This publication has 28 references indexed in Scilit:
- Low-latency Speaker-independent Continuous Speech Separation2019
- Combining Spectral and Spatial Features for Deep Learning Based Blind Speaker SeparationIEEE/ACM Transactions on Audio, Speech, and Language Processing, 2018
- Recognizing Overlapped Speech in Meetings: A Multichannel Separation Approach Using Neural Networks2018
- Multi-Microphone Neural Speech Separation for Far-Field Multi-Talker Speech Recognition2018
- Multitalker Speech Separation With Utterance-Level Permutation Invariant Training of Deep Recurrent Neural NetworksIEEE/ACM Transactions on Audio, Speech, and Language Processing, 2017
- Generalization of Multi-Channel Linear Prediction Methods for Blind MIMO Impulse Response ShorteningIEEE Transactions on Audio, Speech, and Language Processing, 2012
- Generating sensor signals in isotropic noise fieldsThe Journal of the Acoustical Society of America, 2007
- Acoustic Beamforming for Speaker Diarization of MeetingsIEEE Transactions on Audio, Speech, and Language Processing, 2007
- The AMI Meeting Corpus: A Pre-announcement2006
- Image method for efficiently simulating small-room acousticsThe Journal of the Acoustical Society of America, 1979