arXiv subjects

Zexu Pan

Publications and source records attributed to Zexu Pan.

At least 19 recordsLinked to original sources

arXiv2026

TF-MossFormer: Integrating Convolution Gated Local-Global Attentions for Enhanced Time-Frequency Domain Monaural Speech Separation

Transformers with global attention capture long-range dependencies but can miss the fine-grained local continuity crucial for speech separation. We propose TF-MossFormer, a time-frequency transformer that combines local and global attention to jointly model short- and long-range contexts for monaural speech separation. At its core is a content-aware sliding-window attention mechanism that dynamically adapts receptive fields for stronger local interactions, avoiding the rigidity of static convolutions. Unlike time-domain chunk-based methods, TF-MossFormer leverages the 2D spectrogram to model structure along both time and frequency axes. Convolutional gating between attention layers further improves feature selection and information flow. TF-MossFormer achieves SI-SDRi of 22.6, 24.0, and 24.4 dB on WSJ0-2Mix with 5.9M, 16.9M, and 25.4M parameters, respectively, outperforming prior approaches.

cs.SD↗

arXiv2026

SelectTSL: Prompt-Guided Selective Target Sound Localization in Complex Scenarios

Humans can selectively attend to a target sound and estimate its direction in complex scenarios, whereas such selective localization remains challenging for current deep learning-based systems. Sound source localization (SSL) has achieved remarkable success with deep learning, yet most methods localize all active sources without selectivity. Conversely, target sound extraction (TSE) extracts sources using multimodal prompts but typically fails to preserve the multichannel spatial information required for accurate localization. To bridge this gap, we formulate the task of prompt-guided selective target sound localization and propose SelectTSL, an end-to-end architecture that localizes only the user-specified target in multi-source acoustic scenes. Specifically, we design a target-aware selective localization strategy that employs a Prompt-Guided Selective Attention Module (PGSA) to generate prompt-informed embeddings. These embeddings guide an inter-channel phase difference (IPD) enhancer to refine raw phase cues, fusing with target magnitudes to jointly estimate direction of arrival (DoA) and target-source cardinality, i.e., the number of target sound sources. This coupled design effectively focuses on the user-specified target spatial cues for selective localization and also handles time-varying numbers of target sources. Extensive experiments on both synthetic data and real-world recordings demonstrate that our proposed method consistently outperforms other baselines and exhibits robust generalization to real acoustic environments.

cs.SD↗

arXiv2026

Beyond Lips: Integrating Gesture and Lip Cues for Robust Audio-visual Speaker Extraction

Most audio-visual speaker extraction methods rely on synchronized lip recording to isolate the speech of a target speaker from a multi-talker mixture. However, in natural human communication, co-speech gestures are also temporally aligned with speech, often emphasizing specific words or syllables. These gestures provide complementary visual cues that can be especially valuable when facial or lip regions are occluded or distant. In this work, we move beyond lip-centric approaches and propose SeLG, a model that integrates both lip and upper-body gesture information for robust speaker extraction. SeLG features a cross-attention-based fusion mechanism that enables each visual modality to query and selectively attend to relevant speech features in the mixture. To improve the alignment of gesture representations with speech dynamics, SeLG also employs a contrastive InfoNCE loss that encourages gesture embeddings to align more closely with corresponding lip embeddings, which are more strongly correlated with speech. Experimental results on the YGD dataset, containing TED talks, demonstrate that the proposed contrastive learning strategy significantly improves gesture-based speaker extraction, and that our proposed SeLG model, by effectively fusing lip and gesture cues with an attention mechanism and InfoNCE loss, achieves superior performance compared to baselines, across both complete and partial (i.e., missing-modality) conditions.

eess.AS↗

arXiv2026

LuSeeL: Language-queried Binaural Universal Sound Event Extraction and Localization

Most universal sound extraction algorithms focus on isolating a target sound event from single-channel audio mixtures. However, the real world is three-dimensional, and binaural audio, which mimics human hearing, can capture richer spatial information, including sound source location. This spatial context is crucial for understanding and modeling complex auditory scenes, as it inherently informs sound detection and extraction. In this work, we propose a language-driven universal sound extraction network that isolates text-described sound events from binaural mixtures by effectively leveraging the spatial cues present in binaural signals. Additionally, we jointly predict the direction of arrival (DoA) of the target sound using spatial features from the extraction network. This dual-task approach exploits complementary location information to improve extraction performance while enabling accurate DoA estimation. Experimental results on the in-the-wild AudioCaps dataset show that our proposed LuSeeL model significantly outperforms single-channel and uni-task baselines.

eess.AS↗

arXiv2026

FlowSE-GRPO: Training Flow Matching Speech Enhancement via Online Reinforcement Learning

Generative speech enhancement offers a promising alternative to traditional discriminative methods by modeling the distribution of clean speech conditioned on noisy inputs. Post-training alignment via reinforcement learning (RL) effectively aligns generative models with human preferences and downstream metrics in domains such as natural language processing, but its use in speech enhancement remains limited, especially for online RL. Prior work explores offline methods like Direct Preference Optimization (DPO); online methods such as Group Relative Policy Optimization (GRPO) remain largely uninvestigated. In this paper, we present the first successful integration of online GRPO into a flow-matching speech enhancement framework, enabling efficient post-training alignment to perceptual and task-oriented metrics with few update steps. Unlike prior GRPO work on Large Language Models, we adapt the algorithm to the continuous, time-series nature of speech and to the dynamics of flow-matching generative models. We show that optimizing a single reward yields rapid metric gains but often induces reward hacking that degrades audio fidelity despite higher scores. To mitigate this, we propose a multi-metric reward optimization strategy that balances competing objectives, substantially reducing overfitting and improving overall performance. Our experiments validate online GRPO for speech enhancement and provide practical guidance for RL-based post-training of generative audio models.

eess.AS↗

arXiv2025

Fun-ASR Technical Report

In recent years, automatic speech recognition (ASR) has witnessed transformative advancements driven by three complementary paradigms: data scaling, model size scaling, and deep integration with large language models (LLMs). However, LLMs are prone to hallucination, which can significantly degrade user experience in real-world ASR applications. In this paper, we present Fun-ASR, a large-scale, LLM-based ASR system that synergistically combines massive data, large model capacity, LLM integration, and reinforcement learning to achieve state-of-the-art performance across diverse and complex speech recognition scenarios. Moreover, Fun-ASR is specifically optimized for practical deployment, with enhancements in streaming capability, noise robustness, code-switching, hotword customization, and satisfying other real-world application requirements. Experimental results show that while most LLM-based ASR systems achieve strong performance on open-source benchmarks, they often underperform on real industry evaluation sets. Thanks to production-oriented optimizations, Fun-ASR achieves state-of-the-art performance on real application datasets, demonstrating its effectiveness and robustness in practical settings. The code and models are accessible at https://github.com/FunAudioLLM/Fun-ASR .

cs.CL↗

arXiv2025

MeMo: Attentional Momentum for Real-Time Audio-Visual Target Speaker Extraction Under Impaired Visual Conditions

Audio-visual Target Speaker Extraction (AV-TSE) aims to isolate a target speaker's voice from multi-speaker environments by leveraging visual cues as guidance. However, the performance of AV-TSE systems heavily relies on the quality of these visual cues. In extreme scenarios where visual cues are missing or severely degraded, the system may fail to accurately extract the target speaker. In contrast, humans can maintain attention on a target speaker even in the absence of explicit auxiliary information. Motivated by such human cognitive ability, we propose a novel framework called MeMo, which incorporates two adaptive memory banks to store attention-related information. MeMo is specifically designed for real-time scenarios: once initial attention is established, the system maintains attentional momentum over time, even when visual cues become unavailable. We conduct comprehensive experiments to verify the effectiveness of MeMo. Experimental results demonstrate that our proposed framework achieves SI-SNR improvements of at least 2 dB over the corresponding baseline.

cs.SD↗

arXiv2025

ClearerVoice-Studio: Bridging Advanced Speech Processing Research and Practical Deployment

This paper introduces ClearerVoice-Studio, an open-source, AI-powered speech processing toolkit designed to bridge cutting-edge research and practical application. Unlike broad platforms like SpeechBrain and ESPnet, ClearerVoice-Studio focuses on interconnected speech tasks of speech enhancement, separation, super-resolution, and multimodal target speaker extraction. A key advantage is its state-of-the-art pretrained models, including FRCRN with 3 million uses and MossFormer with 2.5 million uses, optimized for real-world scenarios. It also offers model optimization tools, multi-format audio support, the SpeechScore evaluation toolkit, and user-friendly interfaces, catering to researchers, developers, and end-users. Its rapid adoption attracting 3000 GitHub stars and 239 forks highlights its academic and industrial impact. This paper details ClearerVoice-Studio's capabilities, architectures, training strategies, benchmarks, community impact, and future plan. Source code is available at https://github.com/modelscope/ClearerVoice-Studio.

cs.SD↗

arXiv2025

Online Audio-Visual Autoregressive Speaker Extraction

This paper proposes a novel online audio-visual speaker extraction model. In the streaming regime, most studies optimize the audio network only, leaving the visual frontend less explored. We first propose a lightweight visual frontend based on depth-wise separable convolution. Then, we propose a lightweight autoregressive acoustic encoder to serve as the second cue, to actively explore the information in the separated speech signal from past steps. Scenario-wise, for the first time, we study how the algorithm performs when there is a change in focus of attention, i.e., the target speaker. Experimental results on LRS3 datasets show that our visual frontend performs comparably to the previous state-of-the-art on both SkiM and ConvTasNet audio backbones with only 0.1 million network parameters and 2.1 MACs per second of processing. The autoregressive acoustic encoder provides an additional 0.9 dB gain in terms of SI-SNRi, and its momentum is robust against the change in attention.

eess.AS↗

arXiv2025

M3ANet: Multi-scale and Multi-Modal Alignment Network for Brain-Assisted Target Speaker Extraction

The brain-assisted target speaker extraction (TSE) aims to extract the attended speech from mixed speech by utilizing the brain neural activities, for example Electroencephalography (EEG). However, existing models overlook the issue of temporal misalignment between speech and EEG modalities, which hampers TSE performance. In addition, the speech encoder in current models typically uses basic temporal operations (e.g., one-dimensional convolution), which are unable to effectively extract target speaker information. To address these issues, this paper proposes a multi-scale and multi-modal alignment network (M3ANet) for brain-assisted TSE. Specifically, to eliminate the temporal inconsistency between EEG and speech modalities, the modal alignment module that uses a contrastive learning strategy is applied to align the temporal features of both modalities. Additionally, to fully extract speech information, multi-scale convolutions with GroupMamba modules are used as the speech encoder, which scans speech features at each scale from different directions, enabling the model to capture deep sequence information. Experimental results on three publicly available datasets show that the proposed model outperforms current state-of-the-art methods across various evaluation metrics, highlighting the effectiveness of our proposed method. The source code is available at: https://github.com/fchest/M3ANet.

eess.AS↗

arXiv2025

Plug-and-Play Co-Occurring Face Attention for Robust Audio-Visual Speaker Extraction

Audio-visual speaker extraction isolates a target speaker's speech from a mixture speech signal conditioned on a visual cue, typically using the target speaker's face recording. However, in real-world scenarios, other co-occurring faces are often present on-screen, providing valuable speaker activity cues in the scene. In this work, we introduce a plug-and-play inter-speaker attention module to process these flexible numbers of co-occurring faces, allowing for more accurate speaker extraction in complex multi-person environments. We integrate our module into two prominent models: the AV-DPRNN and the state-of-the-art AV-TFGridNet. Extensive experiments on diverse datasets, including the highly overlapped VoxCeleb2 and sparsely overlapped MISP, demonstrate that our approach consistently outperforms baselines. Furthermore, cross-dataset evaluations on LRS2 and LRS3 confirm the robustness and generalizability of our method.

eess.AS↗

arXiv2025

Causal Self-supervised Pretrained Frontend with Predictive Code for Speech Separation

Speech separation (SS) seeks to disentangle a multi-talker speech mixture into single-talker speech streams. Although SS can be generally achieved using offline methods, such a processing paradigm is not suitable for real-time streaming applications. Causal separation models, which rely only on past and present information, offer a promising solution for real-time streaming. However, these models typically suffer from notable performance degradation due to the absence of future context. In this paper, we introduce a novel frontend that is designed to mitigate the mismatch between training and run-time inference by implicitly incorporating future information into causal models through predictive patterns. The pretrained frontend employs a transformer decoder network with a causal convolutional encoder as the backbone and is pretrained in a self-supervised manner with two innovative pretext tasks: autoregressive hybrid prediction and contextual knowledge distillation. These tasks enable the model to capture predictive patterns directly from mixtures in a self-supervised manner. The pretrained frontend subsequently serves as a feature extractor to generate high-quality predictive patterns. Comprehensive evaluations on synthetic and real-world datasets validated the effectiveness of the proposed pretrained frontend.

cs.SD↗

arXiv2025

Context-Aware Two-Step Training Scheme for Domain Invariant Speech Separation

Speech separation seeks to isolate individual speech signals from a multi-talk speech mixture. Despite much progress, a system well-trained on synthetic data often experiences performance degradation on out-of-domain data, such as real-world speech mixtures. To address this, we introduce a novel context-aware, two-stage training scheme for speech separation models. In this training scheme, the conventional end-to-end architecture is replaced with a framework that contains a context extractor and a segregator. The two modules are trained step by step to simulate the speech separation process of an auditory system. We evaluate the proposed training scheme through cross-domain experiments on both synthetic and real-world speech mixtures, and demonstrate that our new scheme effectively boosts separation quality across different domains without adaptation, as measured by signal quality metrics and word error rate (WER). Additionally, an ablation study on the real test set highlights that the context information, including phoneme and word representations from pretrained SSL models, serves as effective domain invariant training targets for separation models.

cs.SD↗

arXiv2025

InspireMusic: Integrating Super Resolution and Large Language Model for High-Fidelity Long-Form Music Generation

We introduce InspireMusic, a framework integrated super resolution and large language model for high-fidelity long-form music generation. A unified framework generates high-fidelity music, songs, and audio, which incorporates an autoregressive transformer with a super-resolution flow-matching model. This framework enables the controllable generation of high-fidelity long-form music at a higher sampling rate from both text and audio prompts. Our model differs from previous approaches, as we utilize an audio tokenizer with one codebook that contains richer semantic information, thereby reducing training costs and enhancing efficiency. This combination enables us to achieve high-quality audio generation with long-form coherence of up to $8$ minutes. Then, an autoregressive transformer model based on Qwen 2.5 predicts audio tokens. Next, we employ a super-resolution flow-matching model to generate high-sampling rate audio with fine-grained details learned from an acoustic codec model. Comprehensive experiments show that the InspireMusic-1.5B-Long model has a comparable performance to recent top-tier open-source systems, including MusicGen and Stable Audio 2.0, on subjective and objective evaluations. The code and pre-trained models are released at https://github.com/FunAudioLLM/InspireMusic.

cs.SD↗

arXiv2025

HiFi-SR: A Unified Generative Transformer-Convolutional Adversarial Network for High-Fidelity Speech Super-Resolution

The application of generative adversarial networks (GANs) has recently advanced speech super-resolution (SR) based on intermediate representations like mel-spectrograms. However, existing SR methods that typically rely on independently trained and concatenated networks may lead to inconsistent representations and poor speech quality, especially in out-of-domain scenarios. In this work, we propose HiFi-SR, a unified network that leverages end-to-end adversarial training to achieve high-fidelity speech super-resolution. Our model features a unified transformer-convolutional generator designed to seamlessly handle both the prediction of latent representations and their conversion into time-domain waveforms. The transformer network serves as a powerful encoder, converting low-resolution mel-spectrograms into latent space representations, while the convolutional network upscales these representations into high-resolution waveforms. To enhance high-frequency fidelity, we incorporate a multi-band, multi-scale time-frequency discriminator, along with a multi-scale mel-reconstruction loss in the adversarial training process. HiFi-SR is versatile, capable of upscaling any input speech signal between 4 kHz and 32 kHz to a 48 kHz sampling rate. Experimental results demonstrate that HiFi-SR significantly outperforms existing speech SR methods across both objective metrics and ABX preference tests, for both in-domain and out-of-domain scenarios (https://github.com/modelscope/ClearerVoice-Studio).

cs.SD↗

arXiv2025

Conditional Latent Diffusion-Based Speech Enhancement Via Dual Context Learning

Recently, the application of diffusion probabilistic models has advanced speech enhancement through generative approaches. However, existing diffusion-based methods have focused on the generation process in high-dimensional waveform or spectral domains, leading to increased generation complexity and slower inference speeds. Additionally, these methods have primarily modelled clean speech distributions, with limited exploration of noise distributions, thereby constraining the discriminative capability of diffusion models for speech enhancement. To address these issues, we propose a novel approach that integrates a conditional latent diffusion model (cLDM) with dual-context learning (DCL). Our method utilizes a variational autoencoder (VAE) to compress mel-spectrograms into a low-dimensional latent space. We then apply cLDM to transform the latent representations of both clean speech and background noise into Gaussian noise by the DCL process, and a parameterized model is trained to reverse this process, conditioned on noisy latent representations and text embeddings. By operating in a lower-dimensional space, the latent representations reduce the complexity of the generation process, while the DCL process enhances the model's ability to handle diverse and unseen noise environments. Our experiments demonstrate the strong performance of the proposed approach compared to existing diffusion-based methods, even with fewer iterative steps, and highlight the superior generalization capability of our models to out-of-domain noise datasets (https://github.com/modelscope/ClearerVoice-Studio).

cs.SD↗

arXiv2025

Improved Feature Extraction Network for Neuro-Oriented Target Speaker Extraction

The recent rapid development of auditory attention decoding (AAD) offers the possibility of using electroencephalography (EEG) as auxiliary information for target speaker extraction. However, effectively modeling long sequences of speech and resolving the identity of the target speaker from EEG signals remains a major challenge. In this paper, an improved feature extraction network (IFENet) is proposed for neuro-oriented target speaker extraction, which mainly consists of a speech encoder with dual-path Mamba and an EEG encoder with Kolmogorov-Arnold Networks (KAN). We propose SpeechBiMamba, which makes use of dual-path Mamba in modeling local and global speech sequences to extract speech features. In addition, we propose EEGKAN to effectively extract EEG features that are closely related to the auditory stimuli and locate the target speaker through the subject's attention information. Experiments on the KUL and AVED datasets show that IFENet outperforms the state-of-the-art model, achieving 36\% and 29\% relative improvements in terms of scale-invariant signal-to-distortion ratio (SI-SDR) under an open evaluation condition.

cs.SD↗

arXiv2025

SSM2Mel: State Space Model to Reconstruct Mel Spectrogram from the EEG

Decoding speech from brain signals is a challenging research problem that holds significant importance for studying speech processing in the brain. Although breakthroughs have been made in reconstructing the mel spectrograms of audio stimuli perceived by subjects at the word or letter level using noninvasive electroencephalography (EEG), there is still a critical gap in precisely reconstructing continuous speech features, especially at the minute level. To address this issue, this paper proposes a State Space Model (SSM) to reconstruct the mel spectrogram of continuous speech from EEG, named SSM2Mel. This model introduces a novel Mamba module to effectively model the long sequence of EEG signals for imagined speech. In the SSM2Mel model, the S4-UNet structure is used to enhance the extraction of local features of EEG signals, and the Embedding Strength Modulator (ESM) module is used to incorporate subject-specific information. Experimental results show that our model achieves a Pearson correlation of 0.069 on the SparrKULee dataset, which is a 38% improvement over the previous baseline.

eess.SP↗