Binaural segregation in multisource reverberant environments

IEEE Trans. Audio Speech Lang. Process.

2010

Abstract-Existing binaural approaches to speech segregation place an exclusive burden on cues related to the location of sound sources in space. These approaches can achieve excellent performance in anechoic conditions but degrade rapidly in realistic environments where room reverberation corrupts localization cues. In this paper, we propose to integrate monaural and binaural processing to achieve segregation and localization of voiced speech in reverberant environments. The proposed approach builds on monaural analysis for simultaneous organization, and combines it with a novel method for generation of location-based cues in a probabilistic framework that jointly achieves localization and sequential organization. We compare localization performance to two existing methods, sequential organization performance to a model-based system that uses only monaural cues, and segregation performance to an exclusively binaural system. Results suggest that the proposed framework allows for improved source localization and robust segregation of voiced speech in environments with considerable reverberation.Index Terms-Binaural speech segregation, computational auditory scene analysis, monaural grouping, sequential organization, sound localization.

mentioning

confidence: 99%

Sequential Organization of Speech in Reverberant Environments by Integrating Monaural Grouping and Binaural Localization

Woodruff

IEEE Trans. Audio Speech Lang. Process.

2010

“…In addition to SNR and ASR, the following abbreviations are used: DOA: direction of arrival HSI: human speech intelligibility HSQ: human speech quality NRR: noise-residual ratio RSR: retained-speech ratio SDR: speech-to-interference ratio Although none of these studies are formulated for real-time implementation, some algorithms are more suitable for real-time operations than others. Of these, the following algorithms that use beamforming to produce T-F masks are most promising: Aarabi and Shi (2004), Roman et al (2006), and Boldt et al (2008). The Aarabi and Shi algorithm based on phase analysis can be viewed as a form of fixed beamforming with given directions of arrival.…”

Section: Related Studiesmentioning

confidence: 99%

“…To deal with the difficulty posed by room reverberation, Roman and Wang (2004) and Roman, Srinivasan, and Wang (2006) proposed using adaptive beamforming to provide the basis for binary T-F masking in order to segregate a target source from a reverberant mixture. This use of beamforming to generate a binary mask should be contrasted with the use of a beamformer to produce a soft mask in conjunction with ICA (see .…”

Section: Beamforming and T-f Maskingmentioning

confidence: 99%

Time-Frequency Masking for Speech Separation and Its Potential for Hearing Aid Design

Trends in Amplification

2008

Self Cite

143

A new approach to the separation of speech from speech-in-noise mixtures is the use of time-frequency (T-F) masking. Originated in the field of computational auditory scene analysis, T-F masking performs separation in the time-frequency domain. This article introduces the T-F masking concept and reviews T-F masking algorithms that separate target speech from either monaural or binaural mixtures, as well as microphone-array recordings. The review emphasizes techniques that are promising for hearing aid design. This article also surveys recent studies that evaluate the perceptual effects of T-F masking techniques, particularly their effectiveness in improving human speech recognition in noise. An assessment is made of the potential benefits of T-F masking methods for the hearing impaired in light of the processing constraints of hearing aids. Finally, several issues pertinent to T-F masking are discussed.

“…Studies show that speech reconstructed from the idea binary mask produces large improvement in human speech intelligibility [3], [13], [35]. Such a goal has been shown to still be reasonable when room reverberation is present [42], [47].…”

mentioning

confidence: 99%

A Supervised Learning Approach to Monaural Segregation of Reverberant Speech

Jin

IEEE Trans. Audio Speech Lang. Process.

2009

Abstract-A major source of signal degradation in real environments is room reverberation. Monaural speech segregation in reverberant environments is a particularly challenging problem. Although inverse filtering has been proposed to partially restore the harmonicity of reverberant speech before segregation, this approach is sensitive to specific source/receiver and room configurations. This paper proposes a supervised learning approach to monaural segregation of reverberant voiced speech, which learns to map from a set of pitch-based auditory features to a grouping cue encoding the posterior probability of a time-frequency (T-F) unit being target dominant given observed features. We devise a novel objective function for the learning process, which directly relates to the goal of maximizing signal-to-noise ratio. The models trained using this objective function yield significantly better T-F unit labeling. A segmentation and grouping framework is utilized to form reliable segments under reverberant conditions and organize them into streams. Systematic evaluations show that our approach produces very promising results under various reverberant conditions and generalizes well to new utterances and new speakers.