Front-end processing for the CHiME-5 dinner party scenario

By Christoph Boeddeker, Jens Heitkaemper, Joerg Schmalenstroeer, Lukas Drude, Jahn Heymann, Reinhold Haeb-Umbach. 2018

146 citations recorded by the source corpus.

This contribution presents a speech enhancement system for the CHiME-5 Dinner Party Scenario. The front-end employs multi-channel linear time-variant filtering and achieves its gains without the use of a neural network. We present an adaptation of blind source separation techniques to the CHiME-5 database which we call Guided Source Separation (GSS). Using the baseline acoustic and language model, the combination of Weighted Prediction Error based dereverberation, guided source separation, and beamforming reduces the WER by 10:54% (relative) for the single array track and by 21:12% (relative) on the multiple array track.

Subfields: Computational Mechanics, Speech & Audio, Signal Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computational Mechanics

In Speech & Audio

In Signal Processing

Citations in the corpus

SpeechStew: Simply Mix All Available Speech Recognition Data to Train One Large Neural Network (2021, positive): “We use guided source separation [49] with 12 channels to enhance our front end for evaluation.”
CHiME-6 Challenge: Tackling Multispeaker Speech Recognition for Unsegmented Recordings (2020, positive): “We provide two baseline speech enhancement front-ends based on open-source implementations of guided source separation (GSS) [39] and BeamformIt [40], respectively.”
Encoder-Decoder Based Attractor Calculation for End-to-End Neural Diarization (2021, methodology): “For example, in guided source separation [4], speech activities are used as constraints to update time-frequency masks of a complex angular central Gaussian mixture model.”
The STC System for the CHiME-6 Challenge (2020, methodology): “So it is possible to use Guided Source Separation (GSS) [12,13], which was developed during the CHiME-5 Challenge [2] and later allowed to improve the recognition accuracy significantly [19, 20].”
Neural Speaker Diarization with Speaker-Wise Chain Rule (2020, background): “Recent studies on the home-party scenario (CHiME-5 [10]) reported that speaker diarization helped improve automatic speech recognition performance [11–13].”
End-to-End Neural Speaker Diarization with Permutation-Free Objectives (2019, background): “It also helps automatic speech recognition performance in multi-speaker conversation scenarios in meetings (ICSI [3,4], AMI [5,6]) and home environments (CHiME5 [6–10]).”
Towards a Competitive End-to-End Speech Recognition for CHiME-6 Dinner Party Transcription (2020, methodology): “We showed that the combination of powerful GSSbased speech enhancement and the use of the RNN-Transducer model is able to achieve competitive results with hybrid systems.”
A Review of Speaker Diarization: Recent Advances with Deep Learning (2021, methodology): “For example, in the CHiME-6 challenge [83], guided source separation (GSS) [95] based multichannel speech extraction techniques have been used to achieve the top result.”
An Investigation into the Effectiveness of Enhancement in ASR Training and Test for Chime-5 Dinner Party Transcription (2019, methodology): “GSS is a spatial mixture model based blind source separation approach which exploits the annotation given in the CHiME-5 database for initialization and, in this way, avoids the frequency permutation problem [14].”