The third ‘CHiME’ speech separation and recognition challenge: Dataset, task and baselines

By Jon Barker, Ricard Marxer, Emmanuel Vincent, Shinji Watanabe. 2015

423 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Evaluation & Benchmarks, 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 Artificial Intelligence

In Evaluation & Benchmarks

In Speech & Audio

In Signal Processing

Citations in the corpus

Deep Speech 2 : End-to-End Speech Recognition in English and Mandarin (2015, methodology): “Finally, we tested our performance on noisy speech using the test sets from the recently completed third CHiME challenge (Barker et al., 2015).”
Light Gated Recurrent Units for Speech Recognition (2018, methodology): “More information on CHiME data can be found in [23].”
The DIRHA-ENGLISH corpus and related tasks for distant-speech recognition in domestic environments (2015, background): “Along this direction, strong efforts have been spent recently by the international scientiﬁc community, through the development of corpora and challenges, such as REVERB [15], CHIME [16, 17] and ASpIRE.”
Batch-normalized joint training for DNN-based distant speech recognition (2016, methodology): “The WSJ experiments are based on the popular wsj5k task (aligned with the CHiME 3 [6] task) and are conducted under two different acoustic conditions.”
VoiceHome-2, an extended corpus for multichannel speech processing in real homes (2019, neutral): “…environments and application scenarios such as voice command for cars (Aurora-3, 2000; Hansen et al., 2001; Lee et al., 2004) and in public spaces (Barker et al., 2015), automatic transcription of lectures (Lamel et al., 1994), meetings (Janin et al., 2003; Mostefa et al., 2007; Renals et al.,…”
Realistic Multi-Microphone Data Simulation for Distant Speech Recognition (2016, background): “Valuable examples are the corpora released under the ChiME [24, 25] and REVERB [26] challenges, which have contributed to deﬁne common tasks, baselines and evaluation frameworks across researchers.”
The Pytorch-kaldi Speech Recognition Toolkit (2018, methodology): “The CHiME experiments were based on the single channel setting [7].”
Speaker-Invariant Training Via Adversarial Learning (2018, methodology): “The CHiME-3 dataset is released with the 3rd CHiME speech Separation and Recognition Challenge [27], which incorporates the Wall Street Journal corpus sentences spoken in challenging noisy environments, recorded using a 6-channel tablet based microphone array.”
FaSNet: Low-Latency Adaptive Beamforming for Multi-Microphone Audio Processing (2019, methodology): “Multichannel noisy ASR : We use the 3rd CHiME challenge dataset [38] for ASR task.”
Student-teacher network learning with enhanced features (2017, background): “[6] T. Yoshioka, N. Ito, M. Delcroix, A. Ogawa, K. Kinoshita, M. Fujimoto, C. Yu, W. J. Fabian, M. Espi, T. Higuchi et al., “The NTT CHiME-3 system: Advances in speech enhancement and recognition for mobile multi-microphone devices,” in Proceedings of IEEE Workshop on Automatic Speech Recognition an”
Deep Learning for Environmentally Robust Speech Recognition (2017, background): “re enhancement approaches. VI. MULTI-CHANNEL TECHNIQUES Microphone arrays and multi-channel processing techniques have recently played an increasingly signiﬁcant role in the development of robust ASR [8], [9]. A central approach is acoustical beamforming, i.e., spatio-temporal ﬁltering that operates”
An Investigation into the Effectiveness of Enhancement in ASR Training and Test for Chime-5 Dinner Party Transcription (2019, background): “However, there has been a long debate whether it is advisable to apply speech enhancement on data used for ASR training, because it is generally agreed upon that the recognizer should be exposed to as much acoustic variability as possible during training, as long as this variability matches the test”
Sound Event Localization and Detection of Overlapping Sources Using Convolutional Recurrent Neural Networks (2018, methodology): “The IR estimation tool [61] was applied independently on all 32 channels of the Eigenmike.”
Twin Regularization for online speech recognition (2018, methodology): “Additional experiments were conducted with the CHiME 4 dataset [34], that is based on speech data recorded in four noisy environments (on a bus, cafe, pedestrian area, and street junction).”
An analysis of environment, microphone and data simulation mismatches in robust speech recognition (2017, neutral): “For more details, see Barker et al. (2015).”
The Automatic Speech recogition In Reverberant Environments (ASpIRE) challenge (2015, neutral): “The 3rd CHiME challenge [14] added new variability to the distant microphone speech recognition task by using a mobile tablet device to capture live speech with an array of six microphones positioned around the tablet frame in noisy environments rather than by using remixing as in the previous two c”
Returnn: The RWTH extensible training framework for universal recurrent neural networks (2016, methodology): “We demonstrate the performance of RETURNN on frame-wise labeled speech data from the CHiME dataset [36].”
Open Graph Benchmark: Datasets for Machine Learning on Graphs (2020, neutral): “…by I MAGE N ET (Deng et al., 2009) and MS COCO (Lin et al., 2014) in computer vision, GLUE B ENCHMARK (Wang et al., 2018) and SQ U AD (Rajpurkar et al., 2016) in natural language processing, and L IBRI S PEECH (Panayotov et al., 2015) and CH I ME (Barker et al., 2015) in speech processing.”
Speech Enhancement and Dereverberation With Diffusion-Based Generative Models (2022, positive): “a) WSJ0-CHiME3: We create the WSJ0-CHiME3 dataset using clean speech utterances from the Wall Street Journal (WSJ0) dataset [49] and noise signals from the CHiME3 dataset [50].”