The CHiME challenge series aims to advance far field speech recognition technology by promoting research at the interface of signal processing and automatic speech recognition. This paper presents the design and outcomes of the 3rd CHiME Challenge, which targets the performance of automatic speech recognition in a real-world, commercially-motivated scenario: a person talking to a tablet device that has been fitted with a six-channel microphone array. The paper describes the data collection, the task definition and the baseline systems for data simulation, enhancement and recognition. The paper then presents an overview of the 26 systems that were submitted to the challenge focusing on the strategies that proved to be most successful relative to the MVDR array processing and DNN acoustic modeling reference system. Challenge findings related to the role of simulated data in system training and evaluation are discussed.

This paper presents a bi-view (front and side) audiovisual Lombard speech corpus, which is freely available for download. It contains 5400 utterances (2700 Lombard and 2700 plain reference utterances), produced by 54 talkers, with each utterance in the dataset following the same sentence format as the audiovisual “Grid” corpus [Cooke, Barker, Cunningham, and Shao (2006). J. Acoust. Soc. Am. 120(5), 2421–2424]. Analysis of this dataset confirms previous research, showing prominent acoustic, phonetic, and articulatory speech modifications in Lombard speech. In addition, gender differences are observed in the size of Lombard effect. Specifically, female talkers exhibit a greater increase in estimated vowel duration and a greater reduction in F2 frequency.

The study of non-human animals’ communication systems generally relies on the transcription of vocal sequences using a finite set of discrete units. This set is referred to as a vocal repertoire, which is specific to a species or a sub-group of a species. When conducted by human experts, the formal description of vocal repertoires can be laborious and/or biased. This motivates computerised assistance for this procedure, for which machine learning algorithms represent a good opportunity. Unsupervised clustering algorithms are suited for grouping close points together, provided a relevant representation. This paper therefore studies a new method for encoding vocalisations, allowing for automatic clustering to alleviate vocal repertoire characterisation. Borrowing from deep representation learning, we use a convolutional auto-encoder network to learn an abstract representation of vocalisations. We report on the quality of the learnt representation, as well as of state of the art methods, by quantifying their agreement with expert labelled vocalisation types from 8 datasets of other studies across 6 species (birds and marine mammals). With this benchmark, we demonstrate that using auto-encoders improves the relevance of vocalisation representation which serves repertoire characterisation using a very limited number of settings. We also publish a Python package for the bioacoustic community to train their own vocalisation auto-encoders or use a pretrained encoder to browse vocal repertoires and ease unit wise annotation.

The extraction of pitch information is arguably one of the most important tasks in automatic music description systems. However, previous research and evaluation datasets dealing with pitch estimation focused on relatively limited kinds of musical data. This work aims to broaden this scope by addressing symphonic western classical music recordings, focusing on pitch estimation for melody extraction. This material is characterized by a high number of overlapping sources, and by the fact that the melody may be played by different instrumental sections, often alternating within an excerpt. We evaluate the performance of eleven state-of-the-art pitch salience functions, multipitch estimation and melody extraction algorithms when determining the sequence of pitches corresponding to the main melody in a varied set of pieces. An important contribution of the present study is the proposed evaluation framework, including the annotation methodology, generated dataset and evaluation metrics. The results show that the assumptions made by certain methods hold better than others when dealing with this type of music signal, leading to a better performance. Additionally, we propose a simple method for combining the output of several algorithms, with promising results.