Andong Li

Cunhang Fan, Enrui Liu, Jing Zhou et al.

Although artificial neural network (ANN) based speech enhancement (SE) methods demonstrate excellent performance, the high computational complexity and high energy consumption hinder their deployment in practical front-end processing tasks.} Currently, the spiking neural networks (SNNs) have shown potential in reducing power consumption. However, the discrete binary activation and complex spatio-temporal dynamics of SNNs often result in information loss. The current challenge therefore focuses on how to maintain performance and reduce computational complexity. To address this issue, this work propose a Dual-Branch Hybrid Neural (DBHN) Network. 1) In terms of network architecture: A dual-branch network integrating ANN and SNN was designed, where the SNN branch reduces power consumption while the ANN branch addresses information loss; The BandSplit and Time-Frequency (TF) -Mamba modules were developed to simultaneously compress energy consumption and enhance model performance; Spiking Feature Extraction Group (SFEG) and Information Transformation Block (ITB) components were implemented with residual connections to mitigate information loss while further refining feature representations. 2) To facilitate inter-branch information fusion: An Interaction module was designed to promote information exchange at various stages of the dual-branch network; A TF-Cross Attention-Fusion module was designed to perform time-frequency domain fusion of dual-branch information while data-adaptively guiding the SNN branch to retain more critical information. Results show that the proposed model maintains superior performance across three public datasets while achieving an average 7.5 fold reduction in computational complexity compared to baseline models.

Chengzhong Wang, Andong Li, Dingding Yao et al.

While deep learning has advanced speech enhancement (SE), effective phase modeling remains challenging, as conventional networks typically operate within a flat Euclidean feature space, which is not easy to model the underlying circular topology of the phase. To address this, we propose a magnitude-phase dual-stream framework that aligns the phase stream with its intrinsic circular geometry by enforcing Global Rotation Equivariance (GRE) characteristic. Specifically, we introduce a Magnitude-Phase Interactive Convolutional Module (MPICM) for modulus-based information exchange and a Hybrid-Attention Dual Feed-Forward Network (HADF) bottleneck for unified feature fusion, both of which are designed to preserve GRE in the phase stream. Comprehensive evaluations are conducted across phase retrieval, denoising, dereverberation, and bandwidth extension tasks to validate the superiority of the proposed method over multiple advanced baselines. Notably, the proposed architecture reduces Phase Distance by over 20\% in the phase retrieval task and improves PESQ by more than 0.1 in zero-shot cross-corpus denoising evaluations. The overall superiority is also established in universal SE tasks involving mixed distortions. Qualitative analysis further reveals that the learned phase features exhibit distinct periodic patterns, which are consistent with the intrinsic circular nature of the phase. The source code is available at https://github.com/wangchengzhong/GRE-Net.

Yukun Chen, Zhaoxi Mu, Andong Li et al.

Despite the remarkable progress in the synthesis speed and fidelity of neural vocoders, their high energy consumption remains a critical barrier to practical deployment on computationally restricted edge devices. Spiking Neural Networks (SNNs), widely recognized for their high energy efficiency due to their event-driven nature, offer a promising solution for low-resource scenarios. In this paper, we propose Spiking Vocos, a novel spiking neural vocoder with ultra-low energy consumption, built upon the efficient Vocos framework. To mitigate the inherent information bottleneck in SNNs, we design a Spiking ConvNeXt module to reduce Multiply-Accumulate (MAC) operations and incorporate an amplitude shortcut path to preserve crucial signal dynamics. Furthermore, to bridge the performance gap with its Artificial Neural Network (ANN) counterpart, we introduce a self-architectural distillation strategy to effectively transfer knowledge. A lightweight Temporal Shift Module is also integrated to enhance the model's ability to fuse information across the temporal dimension with negligible computational overhead. Experiments demonstrate that our model achieves performance comparable to its ANN counterpart, with UTMOS and PESQ scores of 3.74 and 3.45 respectively, while consuming only 14.7% of the energy. The source code is available at https://github.com/pymaster17/Spiking-Vocos.

Yifan Liang, Fangkun Liu, Andong Li et al.

Recent advancements in visual speech recognition (VSR) have promoted progress in lip-to-speech synthesis, where pre-trained VSR models enhance the intelligibility of synthesized speech by providing valuable semantic information. The success achieved by cascade frameworks, which combine pseudo-VSR with pseudo-text-to-speech (TTS) or implicitly utilize the transcribed text, highlights the benefits of leveraging VSR models. However, these methods typically rely on mel-spectrograms as an intermediate representation, which may introduce a key bottleneck: the domain gap between synthetic mel-spectrograms, generated from inherently error-prone lip-to-speech mappings, and real mel-spectrograms used to train vocoders. This mismatch inevitably degrades synthesis quality. To bridge this gap, we propose Natural Lip-to-Speech (NaturalL2S), an end-to-end framework integrating acoustic inductive biases with differentiable speech generation components. Specifically, we introduce a fundamental frequency (F0) predictor to capture prosodic variations in synthesized speech. The predicted F0 then drives a Differentiable Digital Signal Processing (DDSP) synthesizer to generate a coarse signal which serves as prior information for subsequent speech synthesis. Additionally, instead of relying on a reference speaker embedding as an auxiliary input, our approach achieves satisfactory performance on speaker similarity without explicitly modelling speaker characteristics. Both objective and subjective evaluation results demonstrate that NaturalL2S can effectively enhance the quality of the synthesized speech when compared to state-of-the-art methods. Our demonstration page is accessible at https://yifan-liang.github.io/NaturalL2S/.

Guochen Yu, Andong Li, Hui Wang et al.

The decoupling-style concept begins to ignite in the speech enhancement area, which decouples the original complex spectrum estimation task into multiple easier sub-tasks i.e., magnitude-only recovery and the residual complex spectrum estimation)}, resulting in better performance and easier interpretability. In this paper, we propose a dual-branch federative magnitude and phase estimation framework, dubbed DBT-Net, for monaural speech enhancement, aiming at recovering the coarse- and fine-grained regions of the overall spectrum in parallel. From the complementary perspective, the magnitude estimation branch is designed to filter out dominant noise components in the magnitude domain, while the complex spectrum purification branch is elaborately designed to inpaint the missing spectral details and implicitly estimate the phase information in the complex-valued spectral domain. To facilitate the information flow between each branch, interaction modules are introduced to leverage features learned from one branch, so as to suppress the undesired parts and recover the missing components of the other branch. Instead of adopting the conventional RNNs and temporal convolutional networks for sequence modeling, we employ a novel attention-in-attention transformer-based network within each branch for better feature learning. More specially, it is composed of several adaptive spectro-temporal attention transformer-based modules and an adaptive hierarchical attention module, aiming to capture long-term time-frequency dependencies and further aggregate intermediate hierarchical contextual information. Comprehensive evaluations on the WSJ0-SI84 + DNS-Challenge and VoiceBank + DEMAND dataset demonstrate that the proposed approach consistently outperforms previous advanced systems and yields state-of-the-art performance in terms of speech quality and intelligibility.

Chengshi Zheng, Wenzhe Liu, Andong Li et al.

It is highly desirable that speech enhancement algorithms can achieve good performance while keeping low latency for many applications, such as digital hearing aids, acoustically transparent hearing devices, and public address systems. To improve the performance of traditional low-latency speech enhancement algorithms, a deep filter-bank equalizer (FBE) framework was proposed, which integrated a deep learning-based subband noise reduction network with a deep learning-based shortened digital filter mapping network. In the first network, a deep learning model was trained with a controllable small frame shift to satisfy the low-latency demand, i.e., $\le$ 4 ms, so as to obtain (complex) subband gains, which could be regarded as an adaptive digital filter in each frame. In the second network, to reduce the latency, this adaptive digital filter was implicitly shortened by a deep learning-based framework, and was then applied to noisy speech to reconstruct the enhanced speech without the overlap-add method. Experimental results on the WSJ0-SI84 corpus indicated that the proposed deep FBE with only 4-ms latency achieved much better performance than traditional low-latency speech enhancement algorithms in terms of the indices such as PESQ, STOI, and the amount of noise reduction.

Wenzhe Liu, Andong Li, Chengshi Zheng et al.

Most deep learning-based multi-channel speech enhancement methods focus on designing a set of beamforming coefficients to directly filter the low signal-to-noise ratio signals received by microphones, which hinders the performance of these approaches. To handle these problems, this paper designs a causal neural beam filter that fully exploits the spatial-spectral information in the beam domain. Specifically, multiple beams are designed to steer towards all directions using a parameterized super-directive beamformer in the first stage. After that, the neural spatial filter is learned by simultaneously modeling the spatial and spectral discriminability of the speech and the interference, so as to extract the desired speech coarsely in the second stage. Finally, to further suppress the interference components especially at low frequencies, a residual estimation module is adopted to refine the output of the second stage. Experimental results demonstrate that the proposed approach outperforms many state-of-the-art multi-channel methods on the generated multi-channel speech dataset based on the DNS-Challenge dataset.

Linjuan Cheng, Chengshi Zheng, Andong Li et al.

In hands-free communication system, the coupling between loudspeaker and microphone generates echo signal, which can severely influence the quality of communication. Meanwhile, various types of noise in communication environments further reduce speech quality and intelligibility. It is difficult to extract the near-end signal from the microphone signal within one step, especially in low signal-to-noise ratio scenarios. In this paper, we propose a deep complex network approach to address this issue. Specially, we decompose the stereophonic acoustic echo cancellation into two stages, including linear stereophonic acoustic echo cancellation module and residual echo suppression module, where both modules are based on deep learning architectures. A multi-frame filtering strategy is introduced to benefit the estimation of linear echo by capturing more inter-frame information. Moreover, we decouple the complex spectral mapping into magnitude estimation and complex spectrum refinement. Experimental results demonstrate that our proposed approach achieves stage-of-the-art performance over previous advanced algorithms under various conditions.

Andong Li, Zhaohong Deng, Qiongdan Lou et al.

In clinical practice, electroencephalography (EEG) plays an important role in the diagnosis of epilepsy. EEG-based computer-aided diagnosis of epilepsy can greatly improve the ac-curacy of epilepsy detection while reducing the workload of physicians. However, there are many challenges in practical applications for personalized epileptic EEG detection (i.e., training of detection model for a specific person), including the difficulty in extracting effective features from one single view, the undesirable but common scenario of lacking sufficient training data in practice, and the no guarantee of identically distributed training and test data. To solve these problems, we propose a TSK fuzzy system-based epilepsy detection algorithm that integrates multi-view collaborative transfer learning. To address the challenge due to the limitation of single-view features, multi-view learning ensures the diversity of features by extracting them from different views. The lack of training data for building a personalized detection model is tackled by leveraging the knowledge from the source domain (reference scene) to enhance the performance of the target domain (current scene of interest), where mismatch of data distributions between the two domains is resolved with adaption technique based on maximum mean discrepancy. Notably, the transfer learning and multi-view feature extraction are performed at the same time. Furthermore, the fuzzy rules of the TSK fuzzy system equip the model with strong fuzzy logic inference capability. Hence, the proposed method has the potential to detect epileptic EEG signals effectively, which is demonstrated with the positive results from a large number of experiments on the CHB-MIT dataset.

Guochen Yu, Andong Li, Chengshi Zheng et al.

Curriculum learning begins to thrive in the speech enhancement area, which decouples the original spectrum estimation task into multiple easier sub-tasks to achieve better performance. Motivated by that, we propose a dual-branch attention-in-attention transformer dubbed DB-AIAT to handle both coarse- and fine-grained regions of the spectrum in parallel. From a complementary perspective, a magnitude masking branch is proposed to coarsely estimate the overall magnitude spectrum, and simultaneously a complex refining branch is elaborately designed to compensate for the missing spectral details and implicitly derive phase information. Within each branch, we propose a novel attention-in-attention transformer-based module to replace the conventional RNNs and temporal convolutional networks for temporal sequence modeling. Specifically, the proposed attention-in-attention transformer consists of adaptive temporal-frequency attention transformer blocks and an adaptive hierarchical attention module, aiming to capture long-term temporal-frequency dependencies and further aggregate global hierarchical contextual information. Experimental results on Voice Bank + DEMAND demonstrate that DB-AIAT yields state-of-the-art performance (e.g., 3.31 PESQ, 95.6% STOI and 10.79dB SSNR) over previous advanced systems with a relatively small model size (2.81M).

Guochen Yu, Andong Li, Yutian Wang et al.

For the lack of adequate paired noisy-clean speech corpus in many real scenarios, non-parallel training is a promising task for DNN-based speech enhancement methods. However, because of the severe mismatch between input and target speeches, many previous studies only focus on the magnitude spectrum estimation and remain the phase unaltered, resulting in the degraded speech quality under low signal-to-noise ratio conditions. To tackle this problem, we decouple the difficult target w.r.t. original spectrum optimization into spectral magnitude and phase, and a novel Cycle-in-Cycle generative adversarial network (dubbed CinCGAN) is proposed to jointly estimate the spectral magnitude and phase information stage by stage under unpaired data. In the first stage, we pretrain a magnitude CycleGAN to coarsely estimate the spectral magnitude of clean speech. In the second stage, we incorporate the pretrained CycleGAN with a complex-valued CycleGAN as a cycle-in-cycle structure to simultaneously recover phase information and refine the overall spectrum. Experimental results demonstrate that the proposed approach significantly outperforms previous baselines under non-parallel training. The evaluation on training the models with standard paired data also shows that CinCGAN achieves remarkable performance especially in reducing background noise and speech distortion.

Andong Li, Wenzhe Liu, Chengshi Zheng et al.

The spatial covariance matrix has been considered to be significant for beamformers. Standing upon the intersection of traditional beamformers and deep neural networks, we propose a causal neural beamformer paradigm called Embedding and Beamforming, and two core modules are designed accordingly, namely EM and BM. For EM, instead of estimating spatial covariance matrix explicitly, the 3-D embedding tensor is learned with the network, where both spectral and spatial discriminative information can be represented. For BM, a network is directly leveraged to derive the beamforming weights so as to implement filter-and-sum operation. To further improve the speech quality, a post-processing module is introduced to further suppress the residual noise. Based on the DNS-Challenge dataset, we conduct the experiments for multichannel speech enhancement and the results show that the proposed system outperforms previous advanced baselines by a large margin in multiple evaluation metrics.

Andong Li, Wenzhe Liu, Xiaoxue Luo et al.

Background noise and room reverberation are regarded as two major factors to degrade the subjective speech quality. In this paper, we propose an integrated framework to address simultaneous denoising and dereverberation under complicated scenario environments. It adopts a chain optimization strategy and designs four sub-stages accordingly. In the first two stages, we decouple the multi-task learning w.r.t. complex spectrum into magnitude and phase, and only implement noise and reverberation removal in the magnitude domain. Based on the estimated priors above, we further polish the spectrum in the third stage, where both magnitude and phase information are explicitly repaired with the residual learning. Due to the data mismatch and nonlinear effect of DNNs, the residual noise often exists in the DNN-processed spectrum. To resolve the problem, we adopt a light-weight algorithm as the post-processing module to capture and suppress the residual noise in the non-active regions. In the Interspeech 2021 Deep Noise Suppression (DNS) Challenge, our submitted system ranked top-1 for the real-time track in terms of Mean Opinion Score (MOS) with ITU-T P.835 framework

Andong Li, Chengshi Zheng, Lu Zhang et al.

The capability of the human to pay attention to both coarse and fine-grained regions has been applied to computer vision tasks. Motivated by that, we propose a collaborative learning framework in the complex domain for monaural noise suppression. The proposed system consists of two principal modules, namely spectral feature extraction module (FEM) and stacked glance-gaze modules (GGMs). In FEM, the UNet-block is introduced after each convolution layer, enabling the feature recalibration from multiple scales. In each GGM, we decompose the multi-target optimization in the complex spectrum into two sub-tasks. Specifically, the glance path aims to suppress the noise in the magnitude domain to obtain a coarse estimation, and meanwhile, the gaze path attempts to compensate for the lost spectral detail in the complex domain. The two paths work collaboratively and facilitate spectral estimation from complementary perspectives. Besides, by repeatedly unfolding the GGMs, the intermediate result can be iteratively refined across stages and lead to the ultimate estimation of the spectrum. The experiments are conducted on the WSJ0-SI84, DNS-Challenge dataset, and Voicebank+Demand dataset. Results show that the proposed approach achieves state-of-the-art performance over previous advanced systems on the WSJ0-SI84 and DNS-Challenge dataset, and meanwhile, competitive performance is achieved on the Voicebank+Demand corpus.

Andong Li, Chengshi Zheng, Lu Zhang et al.

In real scenarios, it is often necessary and significant to control the inference speed of speech enhancement systems under different conditions. To this end, we propose a stage-wise adaptive inference approach with early exit mechanism for progressive speech enhancement. Specifically, in each stage, once the spectral distance between adjacent stages lowers the empirically preset threshold, the inference will terminate and output the estimation, which can effectively accelerate the inference speed. To further improve the performance of existing speech enhancement systems, PL-CRN++ is proposed, which is an improved version over our preliminary work PL-CRN and combines stage recurrent mechanism and complex spectral mapping. Extensive experiments are conducted on the TIMIT corpus, the results demonstrate the superiority of our system over state-of-the-art baselines in terms of PESQ, ESTOI and DNSMOS. Moreover, by adjusting the threshold, we can easily control the inference efficiency while sustaining the system performance.

Andong Li, Wenzhe Liu, Xiaoxue Luo et al.

It remains a tough challenge to recover the speech signals contaminated by various noises under real acoustic environments. To this end, we propose a novel system for denoising in the complicated applications, which is mainly comprised of two pipelines, namely a two-stage network and a post-processing module. The first pipeline is proposed to decouple the optimization problem w:r:t: magnitude and phase, i.e., only the magnitude is estimated in the first stage and both of them are further refined in the second stage. The second pipeline aims to further suppress the remaining unnatural distorted noise, which is demonstrated to sufficiently improve the subjective quality. In the ICASSP 2021 Deep Noise Suppression (DNS) Challenge, our submitted system ranked top-1 for the real-time track 1 in terms of Mean Opinion Score (MOS) with ITU-T P.808 framework.

Andong Li, Chengshi Zheng, Renhua Peng et al.

In low signal-to-noise ratio conditions, it is difficult to effectively recover the magnitude and phase information simultaneously. To address this problem, this paper proposes a two-stage algorithm to decouple the joint optimization problem w.r.t. magnitude and phase into two sub-tasks. In the first stage, only magnitude is optimized, which incorporates noisy phase to obtain a coarse complex clean speech spectrum estimation. In the second stage, both the magnitude and phase components are refined. The experiments are conducted on the WSJ0-SI84 corpus, and the results show that the proposed approach significantly outperforms previous baselines in terms of PESQ, ESTOI, and SDR.

Andong Li, Chengshi Zheng, Renhua Peng et al.

The generative adversarial networks (GANs) have facilitated the development of speech enhancement recently. Nevertheless, the performance advantage is still limited when compared with state-of-the-art models. In this paper, we propose a powerful Dynamic Attention Recursive GAN called DARGAN for noise reduction in the time-frequency domain. Different from previous works, we have several innovations. First, recursive learning, an iterative training protocol, is used in the generator, which consists of multiple steps. By reusing the network in each step, the noise components are progressively reduced in a step-wise manner. Second, the dynamic attention mechanism is deployed, which helps to re-adjust the feature distribution in the noise reduction module. Third, we exploit the deep Griffin-Lim algorithm as the module for phase postprocessing, which facilitates further improvement in speech quality. Experimental results on Voice Bank corpus show that the proposed GAN achieves state-of-the-art performance than previous GAN- and non-GAN-based models

Andong Li, Chengshi Zheng, Renhua Peng et al.

This technical report describes our system that is submitted to the Deep Noise Suppression Challenge and presents the results for the non-real-time track. To refine the estimation results stage by stage, we utilize recursive learning, a type of training protocol which aggravates the information through multiple stages with a memory mechanism. The attention generator network is designed to dynamically control the feature distribution of the noise reduction network. To improve the phase recovery accuracy, we take the complex spectral mapping procedure by decoding both real and imaginary spectra. For the final blind test set, the average MOS improvements of the submitted system in noreverb, reverb, and realrec categories are 0.49, 0.24, and 0.36, respectively.

Andong Li, Chengshi Zheng, Cunhang Fan et al.

A person tends to generate dynamic attention towards speech under complicated environments. Based on this phenomenon, we propose a framework combining dynamic attention and recursive learning together for monaural speech enhancement. Apart from a major noise reduction network, we design a separated sub-network, which adaptively generates the attention distribution to control the information flow throughout the major network. To effectively decrease the number of trainable parameters, recursive learning is introduced, which means that the network is reused for multiple stages, where the intermediate output in each stage is correlated with a memory mechanism. As a result, a more flexible and better estimation can be obtained. We conduct experiments on TIMIT corpus. Experimental results show that the proposed architecture obtains consistently better performance than recent state-of-the-art models in terms of both PESQ and STOI scores.

Andong Li, Chengshi Zheng, Linjuan Cheng et al.

In this paper, we propose a type of neural network with feedback learning in the time domain called FTNet for monaural speech enhancement, where the proposed network consists of three principal components. The first part is called stage recurrent neural network, which is introduced to effectively aggregate the deep feature dependencies across different stages with a memory mechanism and also remove the interference stage by stage. The second part is the convolutional auto-encoder. The third part consists of a series of concatenated gated linear units, which are capable of facilitating the information flow and gradually increasing the receptive fields. Feedback learning is adopted to improve the parameter efficiency and therefore, the number of trainable parameters is effectively reduced without sacrificing its performance. Numerous experiments are conducted on TIMIT corpus and experimental results demonstrate that the proposed network can achieve consistently better performance in terms of both PESQ and STOI scores than two state-of-the-art time domain-based baselines in different conditions.

Andong Li, Chengshi Zheng, Xiaodong Li

For voice communication, it is important to extract the speech from its noisy version without introducing unnaturally artificial noise. By studying the subband mean-squared error (MSE) of the speech for unsupervised speech enhancement approaches and revealing its relationship with the existing loss function for supervised approaches, this paper derives a generalized loss function, when taking the residual noise control into account, for supervised approaches. Our generalized loss function contains the well-known MSE loss function and many other often-used loss functions as special cases. Compared with traditional loss functions, our generalized loss function is more flexible to make a good trade-off between speech distortion and noise reduction. This is because a group of well-studied noise shaping schemes can be introduced to control residual noise for practical applications. Objective and subjective test results verify the importance of residual noise control for the supervised speech enhancement approach.

Andong Li, Minmin Yuan, Chengshi Zheng et al.

Recently, progressive learning has shown its capacity to improve speech quality and speech intelligibility when it is combined with deep neural network (DNN) and long short-term memory (LSTM) based monaural speech enhancement algorithms, especially in low signal-to-noise ratio (SNR) conditions. Nevertheless, due to a large number of parameters and high computational complexity, it is hard to implement in current resource-limited micro-controllers and thus, it is essential to significantly reduce both the number of parameters and the computational load for practical applications. For this purpose, we propose a novel progressive learning framework with causal convolutional recurrent neural networks called PL-CRNN, which takes advantage of both convolutional neural networks and recurrent neural networks to drastically reduce the number of parameters and simultaneously improve speech quality and speech intelligibility. Numerous experiments verify the effectiveness of the proposed PL-CRNN model and indicate that it yields consistent better performance than the PL-DNN and PL-LSTM algorithms and also it gets results close even better than the CRNN in terms of objective measurements. Compared with PL-DNN, PL-LSTM, and CRNN, the proposed PL-CRNN algorithm can reduce the number of parameters up to 93%, 97%, and 92%, respectively.