Continuous Speech Separation: Dataset and Analysis

Abstract

This paper describes a dataset and protocols for evaluating continuous speech separation algorithms. Most prior studies on speech separation use pre-segmented signals of artificially mixed speech utterances which are mostly fully overlapped, and the algorithms are evaluated based on signal-to-distortion ratio or similar performance metrics. However, in natural conversations, a speech signal is continuous, containing both overlapped and overlap-free components. In addition, the signal-based metrics have very weak correlations with automatic speech recognition (ASR) accuracy. We think that not only does this make it hard to assess the practical relevance of the tested algorithms, it also hinders researchers from developing systems that can be readily applied to real scenarios. In this paper, we define continuous speech separation (CSS) as a task of generating a set of non-overlapped speech signals from a continuous audio stream that contains multiple utterances that are partially overlapped by a varying degree. A new real recorded dataset, called LibriCSS, is derived from LibriSpeech by concatenating the corpus utterances to simulate a conversation and capturing the audio replays with far-field microphones. A Kaldi-based ASR evaluation protocol is also established by using a well-trained multi-conditional acoustic model. By using this dataset, several aspects of a recently proposed speakerindependent CSS algorithm are investigated. The dataset and evaluation scripts are available to facilitate the research in this direction.

Speaker Activity Driven Neural Speech Extraction

“…In the second experiment, we used the meeting-like LibriCSS corpus [18], which consists of 8-speaker meeting-like recordings sessions of 10 minutes, obtained by re-recording LibriSpeech utterances played through loudspeakers in a meeting room. The overlap ratio varies from 0 to 40 %.…”

“…We discuss related works in Section 4. In Section 5, we present experimental results based on the LibriCSS corpus [18]. Finally, we conclude the paper in Section 6.…”

Dual-Path Modeling for Long Recording Speech Separation in Meetings

“…It contains 10 hours of audio recordings in regular meeting rooms. Each mini-session 1 in 1 Readers can refer to [23] to get more details.…”

“…All our models in the table use the window size of 2.4s. 0S/L[23]: 0% overlap ratio with short/long silence.…”

MASS: Microphone Array Speech Simulator in Room Acoustic Environment for Multi-Channel Speech Coding and Enhancement

“…However, the microphone array in this dataset is only a circular array and cannot be changed, so it cannot be applied to the scenes that require a specific shape of the microphone array. Chen et al proposed a dataset [14] for evaluating continuous speech separation. In this dataset, the speech signal is continuous, containing both the overlapped and overlap-free components.…”