Deep learning-based stereophonic acoustic echo suppression without decorrelation

Abstract

Traditional stereophonic acoustic echo cancellation algorithms need to estimate acoustic echo paths from stereo loudspeakers to a microphone, which often suffers from the nonuniqueness problem caused by a high correlation between the two far-end signals of these stereo loudspeakers. Many decorrelation methods have already been proposed to mitigate this problem. However, these methods may reduce the audio quality and/or stereophonic spatial perception. This paper proposes to use a convolutional recurrent network (CRN) to suppress the stereophonic echo components by estimating a nonlinear gain, which is then multiplied by the complex spectrum of the microphone signal to obtain the estimated near-end speech without a decorrelation procedure. The CRN includes an encoder-decoder module and two-layer gated recurrent network module, which can take advantage of the feature extraction capability of the convolutional neural networks and temporal modeling capability of recurrent neural networks simultaneously. The magnitude spectra of the two far-end signals are used as input features directly without any decorrelation preprocessing and, thus, both the audio quality and stereophonic spatial perception can be maintained. The experimental results in both the simulated and real acoustic environments show that the proposed algorithm outperforms traditional algorithms such as the normalized least-mean square and Wiener algorithms, especially in situations of low signal-to-echo ratio and high reverberation time RT60.

Kahramanmaraş Merkezli Depremler Sonrası İçin Akademik Öneriler

“…Özellikle, yankı, ortamda bulunan nesneler ve yapıların etkisi, kaynak konumunun belirlenebilme doğruluğunu olumsuz etkileyebilir. Bu faktörlerin etkisini azaltabilmek için oluşturulmuş bir veri kümesi ile eğitilmiş makine öğrenmesi modeli sesin işlenmesinde kullanılabilir (Cheng, Peng, Li, Zheng, & Li, 2021;Haubner & Kellermann, 2022).…”

LCSM: A Lightweight Complex Spectral Mapping Framework for Stereophonic Acoustic Echo Cancellation

“…The classic CRN [10,11] downsample and upsample operations shrink and extend the features along the frequency dimension in the encoder and decoder, respectively. However, for multichannel signal processing tasks, the downsampling could confuse spatial cues making LSTM hard to extract the spatial information, since the frequency information tangle with the channel dimension.…”

“…We utilize three algorithms as baselines for comparison experiments, there is Yang [8] which is a conventional method for addressing the stereo echo suppression problem, and Cheng et al [10] and Zhang [11] are the recent CRN-based methods.…”

“…More recently, deep learning-based methods which intrinsically avoid the non-uniqueness problem have been proposed for solving SAEC, and achieve impressive performance. Cheng et al [10] employs the convolutional recurrent network (CRN) which takes the magnitude spectra of both far-end and microphone signals as inputs to estimate a mask for the near-end magnitude spectrum. In contrast to the mask estimation, Zhang and Wang [11] proposed a method that learn the complex spectral mapping to directly estimate real and imaginary spectra of nearend speech, by using the complex spectrums of far-end and microphone signals as the CRN inputs.…”

A deep complex multi-frame filtering network for stereophonic acoustic echo cancellation

“…Deep learing has not received much attention in SAEC research. A convolutional recurrent network (CRN)-based SAES algorithm has been first proposed in [19], which use deep neural network to directly suppress stereophonic acoustic echo from microphone input signal without decorrelation. And a CRNbased complex SAES with a two-stage approach was proposed in [20].…”