A Linear Neural Network-Based Approach to Stereophonic Acoustic Echo Cancellation

Bekrani, Mehdi; Khong, Andy W. H.; Lotfizad, Mojtaba

doi:10.1109/tasl.2010.2098868

Cited by 9 publications

(6 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…It is very often performed in conjunction with direction of arrival (DoA) estimation, in which only the direction information about the source position is obtained. Both tasks are useful in many practical applications, including increasing the robustness of automatic speech recognition [1] by enhancing the performance of acoustic echo cancellers [2] and autonomous robotics [3], [4]. Despite both DoA and source distance being estimated using multi-channel audio in most practical scenarios, the latter has been largely underresearched [5].…”

Section: Introductionmentioning

confidence: 99%

Speaker Distance Estimation in Enclosures From Single-Channel Audio

Neri,

Politis,

Krause

et al. 2024

IEEE/ACM Trans. Audio Speech Lang. Process.

View full text Add to dashboard Cite

Distance estimation from audio plays a crucial role in various applications, such as acoustic scene analysis, sound source localization, and room modeling. Most studies predominantly center on employing a classification approach, where distances are discretized into distinct categories, enabling smoother model training and achieving higher accuracy but imposing restrictions on the precision of the obtained sound source position. Towards this direction, in this paper we propose a novel approach for continuous distance estimation from audio signals using a convolutional recurrent neural network with an attention module. The attention mechanism enables the model to focus on relevant temporal and spectral features, enhancing its ability to capture fine-grained distance-related information. To evaluate the effectiveness of our proposed method, we conduct extensive experiments using audio recordings in controlled environments with three levels of realism (synthetic room impulse response, measured response with convolved speech, and real recordings) on four datasets (our synthetic dataset, QMULTI-MIT, VoiceHome-2, and STARSS23). Experimental results show that the model achieves an absolute error of 0.11 meters in a noiseless synthetic scenario. Moreover, the results showed an absolute error of about 1.30 meters in the hybrid scenario. The algorithm's performance in the real scenario, where unpredictable environmental factors and noise are prevalent, yields an absolute error of approximately 0.50 meters. For reproducible research purposes we make model, code, and synthetic datasets available at https://github.com/michaelneri/audio-distance-estimation.

show abstract

Section: Introductionmentioning

confidence: 99%

Speaker Distance Estimation in Enclosures From Single-Channel Audio

Neri,

Politis,

Krause

et al. 2024

IEEE/ACM Trans. Audio Speech Lang. Process.

View full text Add to dashboard Cite

show abstract

“…To overcome the effects of the ROE and MPE on the generated soundfield, an improved LS-based algorithm with iteratively estimating the system matrix inverse for computing the loudspeaker array weight coefficients over the zone of interest has been proposed in Chapter 6. The proposed algorithm makes use of a linear network that was originally proposed for the diagonalization of the correlation matrix constructed using a pair of stereophonic input vectors [17] and the estimation of the inverse of a joint-input correlation matrix of input vectors [18] for stereophonic acoustic echo cancellation application. The single-layer linear network is then generalized to a multi-layer network for a better accuracy of matrix inverse estimation.…”

Section: Organization Of the Thesismentioning

confidence: 99%

“…The starting point of the proposed algorithm is the use of a linear network that was originally proposed for the diagonalization of the correlation matrix constructed using a pair of stereophonic input vectors [17] and the estimation of the inverse of a joint-input correlation matrix of input vectors [18] for stereophonic acoustic echo cancellation application.…”

Section: The Proposed Ls-based Iterative Matrix Inverse Algorithmmentioning

confidence: 99%

Section: The Proposed Ls-based Iterative Matrix Inverse Algorithmmentioning

confidence: 99%

“…The techniques proposed in [17] and [18] were originally developed for reducing the interchannel coherence and partial decorrelation of stereophonic signals. In [17] these signals are employed for reducing the eigenvalue spread of input signals for improving the convergence performance of multichannel adaptive filters. In [18] these signals are employed for obtaining a coarse estimation of inverse of joint-input correlation matrix to be used in LMS-Newton adaptive filter in stereophonic acoustic echo cancellation.…”

Section: The Proposed Ls-based Iterative Matrix Inverse Algorithmmentioning

confidence: 99%

See 2 more Smart Citations