The Automatic Speech recogition In Reverberant Environments (ASpIRE) challenge

Abstract

In this paper, we describe the ASpIRE (Automatic Speech recognition In Reverberant Environments) challenge, which asked participants to construct automatic speech recognition systems that were robust to a variety of acoustic environments and recording scenarios without having access to matched training and development data. We discuss the performance of the systems evaluated in the challenge, summarize how those systems were constructed, and draw conclusions about what contributed to the performance levels of the systems on the highly variable, noisy, reverberant evaluation data set constructed for this challenge.

An analysis of environment, microphone and data simulation mismatches in robust speech recognition

“…Research in this field has made great progress thanks to real speech corpora collected for various application scenarios such as voice command for cars (Hansen et al, 2001), smart homes (Ravanelli et al, 2015), or tablets (Barker et al, 2015), and automatic transcription of lectures (Lamel et al, 1994), meetings (Renals et al, 2008), conversations (Harper, 2015), dialogues (Stupakov et al, 2011), game sessions (Fox et al, 2013), or broadcast media (Bell et al, 2015). In most corpora, the training speakers differ from the test speakers.…”

A reassigned front-end for speech recognition

“…Recent speech recognition evaluation campaigns, for example REVERB [16], CHiME-3 [17] and ASpIRE [18], indicate that state-of-the-art systems often choose sophisticated feature extraction methods, such as i-vector and gammatone ceptral coefficient [19], and incorporate additional front-end processing units, such as speech enhancement, beamforming and CS in order to improve recognition performance in real applications.…”

Joint training of DNNs by incorporating an explicit dereverberation structure for distant speech recognition

“…In recent years, substantial progress has been made for distant/reverberant speech recognition by several important challenges, such as REVERB (REverberant Voice Enhancement and Recognition Benchmark) challenge [3], CHiME [4] challenge mainly for solving background noises, and ASpIRE (Automatic Speech Recognition In Reverberant Environments) [5]. Many techniques have been widely investigated, including front-end multichannel and single-channel dereverberation techniques, and back-end acoustic modeling approaches.…”