Multi-channel speech enhancement aims to extract clean speech from a noisy mixture using signals captured from multiple microphones. Recently proposed methods tackle this problem by incorporating deep neural network models with spatial filtering techniques such as the minimum variance distortionless response (MVDR) beamformer. In this paper, we introduce a different research direction by viewing each audio channel as a node lying in a non-Euclidean space and, specifically, a graph. This formulation allows us to apply graph neural networks (GNN) to find spatial correlations among the different channels (nodes). We utilize graph convolution networks (GCN) by incorporating them in the embedding space of a U-Net architecture. We use LibriSpeech dataset and simulate room acoustics data to extensively experiment with our approach using different array types, and number of microphones. Results indicate the superiority of our approach when compared to prior state-of-the-art method.

Reproducing zones of personal sound is a challenging signal processing problem which has garnered considerable research interest in recent years. We introduce in this work an extended method to multizone soundfield reproduction which overcomes issues with speech privacy and quality. Measures of Speech Intelligibility Contrast (SIC) and speech quality are used as cost functions in an optimisation of speech privacy and quality. Novel spatial and (temporal) frequency domain speech masker filter designs are proposed to accompany the optimisation process. Spatial masking filters are designed using multizone soundfield algorithms which are dependent on the target speech multizone reproduction. Combinations of estimates of acoustic contrast and long term average speech spectra are proposed to provide equal masking influence on speech privacy and quality. Spatial aliasing specific to multizone soundfield reproduction geometry is further considered in analytically derived low-pass filters. Simulated and real-world experiments are conducted to verify the performance of the proposed method using semicircular and linear loudspeaker arrays. Simulated implementations of the proposed method show that significant speech intelligibility contrast and speech quality is achievable between zones. A range of Perceptual Evaluation of Speech Quality (PESQ) Mean Opinion Scores (MOS) that indicate good quality are obtained while at the same time providing confidential privacy as indicated by SIC. The simulations also show that the method is robust to variations in the speech, virtual source location, array geometry and number of loudspeakers. Real-world experiments confirm the practicality of the proposed methods by showing that good quality and confidential privacy are achievable.

No abstract

This paper proposes and evaluates an efficient approach for practical reproduction of multizone soundfields for speech sources. The reproduction method, based on a previously proposed approach, utilises weighting parameters to control the soundfield reproduced in each zone whilst minimising the number of loudspeakers required. Proposed here is an interpolation scheme for predicting the weighting parameter values of the multizone soundfield model that otherwise requires significant computational effort. It is shown that initial computation time can be reduced by a factor of 1024 with only 85dB of error in the reproduced soundfield relative to reproduction without interpolated weighting parameters. The perceptual impact on the quality of the speech reproduced using the method is also shown to be negligible. By using pre-saved soundfields determined using the proposed approach, practical reproduction of dynamically weighted multizone soundfields of wideband speech could be achieved in real-time. AN EFFICIENT APPROACH TO DYNAMICALLY WEIGHTED MULTIZONE WIDEBAND REPRODUCTION OF SPEECH SOUNDFIELDS ABSTRACTThis paper proposes and evaluates an efficient approach for practical reproduction of multizone soundfields for speech sources. The reproduction method, based on a previously proposed approach, utilises weighting parameters to control the soundfield reproduced in each zone whilst minimising the number of loudspeakers required. Proposed here is an interpolation scheme for predicting the weighting parameter values of the multizone soundfield model that otherwise requires significant computational effort. It is shown that initial computation time can be reduced by a factor of 1024 with only 85dB  of error in the reproduced soundfield relative to reproduction without interpolated weighting parameters. The perceptual impact on the quality of the speech reproduced using the method is also shown to be negligible. By using pre-saved soundfields determined using the proposed approach, practical reproduction of dynamically weighted multizone soundfields of wideband speech could be achieved in real-time.Index Terms-multizone soundfield reproduction, wideband multizone soundfield, weighted multizone soundfield, look-up tables (LUT), interpolation, sound field synthesis (SFS)

This paper proposes two methods for providing speech privacy between spatial zones in anechoic and reverberant environments. The methods are based on masking the content leaked between regions. The masking is optimised to maximise the speech intelligibility contrast (SIC) between the zones. The first method uses a uniform masker signal that is combined with desired multizone loudspeaker signals and requires acoustic contrast between zones. The second method computes a space-time domain masker signal in parallel with the loudspeaker signals so that the combination of the two emphasises the spectral masking in the targeted quiet zone. Simulations show that it is possible to achieve a significant SIC in anechoic environments whilst maintaining speech quality in the bright zone. ABSTRACTThis paper proposes two methods for providing speech privacy between spatial zones in anechoic and reverberant environments. The methods are based on masking the content leaked between regions. The masking is optimised to maximise the speech intelligibility contrast (SIC) between the zones. The first method uses a uniform masker signal that is combined with desired multizone loudspeaker signals and requires acoustic contrast between zones. The second method computes a space-time domain masker signal in parallel with the loudspeaker signals so that the combination of the two emphasises the spectral masking in the targeted quiet zone. Simulations show that it is possible to achieve a significant SIC in anechoic environments whilst maintaining speech quality in the bright zone.