arXiv subjects

Qiquan Zhang

Publications and source records attributed to Qiquan Zhang.

At least 19 recordsLinked to original sources

arXiv2026

BiEAR: A Human Auditory-Inspired Adaptive Binaural Front-end for Multi-Speaker Localisation and Distance Estimation

We present BiEAR, a human auditory-inspired adaptive binaural front-end for multi-speaker localisation and distance estimation. Inspired by medial olivocochlear (MOC) feedback in human hearing, BiEAR uses a neural controller to adaptively adjust the frequency selectivity of a binaural auditory filterbank during inference. This yields time-frequency adaptive representations for ears, enabling the model to respond to changing acoustic conditions. We evaluate BiEAR on multi-speaker localisation and distance estimation in anechoic and real-room environments. Results show that the adaptive front-end improves localisation accuracy and robustness to unseen speakers and rooms compared with commonly used fixed binaural front-ends. Visualisation and analysis of learned filter adaptations show that BiEAR emphasises informative frequency bands over time. These findings suggest that adaptive, biologically inspired binaural front-ends can improve machine hearing robustness in complex acoustic scenes.

eess.AS↗

arXiv2026

The WER Trap: Shattering the Illusion of Unified Tokens in Speech Language Models

The pursuit of a "unified" discrete token for both speech understanding and generation has led the Speech Language Model (SLM) community to heavily rely on Word Error Rate (WER) -- the core metric for Whisper-style tokenizers -- as the definitive proxy for representation quality. This fosters the assumption that low-WER tokens inherently preserve the information necessary for intelligible acoustic synthesis. We argue this is fundamentally deceptive. While high-frequency tokens succeed in generation tasks due to implicit information leakage, isolating pure semantic information at ultra-low frame rates strips away the finegrained articulation and micro-dynamics essential for ODE-based generation. Empirically validating this requires extreme compression without sacrificing WER -- a methodological bottleneck, as standard fixed-stride downsampling arbitrarily truncates phonetic boundaries. To overcome this, we develop a dynamic compression tokenizer that intelligently aligns representations with semantic boundaries, achieving ultra-low frame rates with exceptionally low WER. Using these isolated "pure" semantic tokens, we expose the WER trap: when conditioning generative models -- even with oracle duration alignments -- the reconstructed speech suffers from severe articulation blur and is rendered acoustically unintelligible. Our findings demonstrate that semantic categorization rewarded by low WER is inherently orthogonal to the continuous phonetic trajectories required for synthesis, shattering the illusion of the unified token and advocating for explicitly decoupled speech representations.

eess.AS↗

arXiv2026

Benchmarking Gaslighting Attacks Against Speech Large Language Models

As Speech Large Language Models (Speech LLMs) become increasingly integrated into voice-based applications, ensuring their robustness against manipulative or adversarial input becomes critical. Although prior work has studied adversarial attacks in text-based LLMs and vision-language models, the unique cognitive and perceptual challenges of speech-based interaction remain underexplored. In contrast, speech presents inherent ambiguity, continuity, and perceptual diversity, which make adversarial attacks more difficult to detect. In this paper, we introduce gaslighting attacks, strategically crafted prompts designed to mislead, override, or distort model reasoning as a means to evaluate the vulnerability of Speech LLMs. Specifically, we construct five manipulation strategies: Anger, Cognitive Disruption, Sarcasm, Implicit, and Professional Negation, designed to test model robustness across varied tasks. It is worth noting that our framework captures both performance degradation and behavioral responses, including unsolicited apologies and refusals, to diagnose different dimensions of susceptibility. Moreover, acoustic perturbation experiments are conducted to assess multi-modal robustness. To quantify model vulnerability, comprehensive evaluation across 5 Speech and multi-modal LLMs on over 10,000 test samples from 5 diverse datasets reveals an average accuracy drop of 24.3% under the five gaslighting attacks, indicating significant behavioral vulnerability. These findings highlight the need for more resilient and trustworthy speech-based AI systems.

cs.CL↗

arXiv2026

Code-switching Speech Recognition Under the Lens: Model- and Data-Centric Perspectives

Code-switching automatic speech recognition (CS-ASR) presents unique challenges due to language confusion introduced by spontaneous intra-sentence switching and accent bias that blurs the phonetic boundaries. Although the constituent languages may be individually high-resource, the scarcity of annotated code-switching data further compounds these challenges. In this paper, we systematically analyze CS-ASR from both model-centric and data-centric perspectives. By comparing state-of-the-art algorithmic methods, including language-specific processing and auxiliary language-aware multi-task learning, we discuss their varying effectiveness across datasets with different linguistic characteristics. On the data side, we first investigate TTS as a data augmentation method. By varying the textual characteristics and speaker accents, we analyze the impact of language confusion and accent bias on CS-ASR. To further mitigate data scarcity and enhance textual diversity, we propose a prompting strategy by simplifying the equivalence constraint theory (SECT) to guide large language models (LLMs) in generating linguistically valid code-switching text. The proposed SECT outperforms existing methods in ASR performance and linguistic quality assessments, generating code-switching text that more closely resembles real-world code-switching text. When used to generate speech-text pairs via TTS, SECT proves effective in improving CS-ASR performance. Our analysis of both model- and data-centric methods underscores that effective CS-ASR requires strategies to be carefully aligned with the specific linguistic characteristics of the code-switching data.

eess.AS↗

arXiv2026

Quantizer-Aware Hierarchical Neural Codec Modeling for Speech Deepfake Detection

Neural audio codecs discretize speech via residual vector quantization (RVQ), forming a coarse-to-fine hierarchy across quantizers. While codec models have been explored for representation learning, their discrete structure remains underutilized in speech deepfake detection. In particular, different quantization levels capture complementary acoustic cues, where early quantizers encode coarse structure and later quantizers refine residual details that reveal synthesis artifacts. Existing systems either rely on continuous encoder features or ignore this quantizer-level hierarchy. We propose a hierarchy-aware representation learning framework that models quantizer-level contributions through learnable global weighting, enabling structured codec representations aligned with forensic cues. Keeping the speech encoder backbone frozen and updating only 4.4% additional parameters, our method achieves relative EER reductions of 46.2% on ASVspoof 2019 and 13.9% on ASVspoof5 over strong baselines.

cs.SD↗

arXiv2026

Adaptive Per-Channel Energy Normalization Front-end for Robust Audio Signal Processing

In audio signal processing, learnable front-ends have shown strong performance across diverse tasks by optimizing task-specific representation. However, their parameters remain fixed once trained, lacking flexibility during inference and limiting robustness under dynamic complex acoustic environments. In this paper, we introduce a novel adaptive paradigm for audio front-ends that replaces static parameterization with a closed-loop neural controller. Specifically, we simplify the learnable front-end LEAF architecture and integrate a neural controller for adaptive representation via dynamically tuning Per-Channel Energy Normalization. The neural controller leverages both the current and the buffered past subband energies to enable input-dependent adaptation during inference. Experimental results on multiple audio classification tasks demonstrate that the proposed adaptive front-end consistently outperforms prior fixed and learnable front-ends under both clean and complex acoustic conditions. These results highlight neural adaptability as a promising direction for the next generation of audio front-ends.

eess.AS↗

arXiv2026

Analytic Incremental Learning For Sound Source Localization With Imbalance Rectification

Sound source localization (SSL) demonstrates remarkable results in controlled settings but struggles in real-world deployment due to dual imbalance challenges: intra-task imbalance arising from long-tailed direction-of-arrival (DoA) distributions, and inter-task imbalance induced by cross-task skews and overlaps. These often lead to catastrophic forgetting, significantly degrading the localization accuracy. To mitigate these issues, we propose a unified framework with two key innovations. Specifically, we design a GCC-PHAT-based data augmentation (GDA) method that leverages peak characteristics to alleviate intra-task distribution skews. We also propose an Analytic dynamic imbalance rectifier (ADIR) with task-adaption regularization, which enables analytic updates that adapt to inter-task dynamics. On the SSLR benchmark, our proposal achieves state-of-the-art (SoTA) results of 89.0% accuracy, 5.3° mean absolute error, and 1.6 backward transfer, demonstrating robustness to evolving imbalances without exemplar storage.

cs.SD↗

arXiv2026

Fun-Audio-Chat Technical Report

Recent advancements in joint speech-text models show great potential for seamless voice interactions. However, existing models face critical challenges: temporal resolution mismatch between speech tokens (25Hz) and text tokens (~3Hz) dilutes semantic information, incurs high computational costs, and causes catastrophic forgetting of text LLM knowledge. We introduce Fun-Audio-Chat, a Large Audio Language Model addressing these limitations via two innovations from our previous work DrVoice. First, Dual-Resolution Speech Representations (DRSR): the Shared LLM processes audio at efficient 5Hz (via token grouping), while the Speech Refined Head generates high-quality tokens at 25Hz, balancing efficiency (~50% GPU reduction) and quality. Second, Core-Cocktail Training, a two-stage fine-tuning with intermediate merging that mitigates catastrophic forgetting. We then apply Multi-Task DPO Training to enhance robustness, audio understanding, instruction-following and voice empathy. This multi-stage post-training enables Fun-Audio-Chat to retain text LLM knowledge while gaining powerful audio understanding, reasoning, and generation. Unlike recent LALMs requiring large-scale audio-text pre-training, Fun-Audio-Chat leverages pre-trained models and extensive post-training. Fun-Audio-Chat 8B and MoE 30B-A3B achieve competitive performance on Speech-to-Text and Speech-to-Speech tasks, ranking top among similar-scale models on Spoken QA benchmarks. They also achieve competitive to superior performance on Audio Understanding, Speech Function Calling, Instruction-Following and Voice Empathy. We develop Fun-Audio-Chat-Duplex, a full-duplex variant with strong performance on Spoken QA and full-duplex interactions. We open-source Fun-Audio-Chat-8B with training and inference code, and provide an interactive demo, at https://github.com/FunAudioLLM/Fun-Audio-Chat .

cs.CL↗

arXiv2025

Learning Time-Graph Frequency Representation for Monaural Speech Enhancement

The Graph Fourier Transform (GFT) has recently demonstrated promising results in speech enhancement. However, existing GFT-based speech enhancement approaches often employ fixed graph topologies to build the graph Fourier basis, whose the representation lacks the adaptively and flexibility. In addition, they suffer from the numerical errors and instability introduced by matrix inversion in GFT based on both Singular Value Decomposition (GFT-SVD) and Eigen Vector Decomposition (GFT-EVD). Motivated by these limitations, this paper propose a simple yet effective learnable GFT-SVD framework for speech enhancement. Specifically, we leverage graph shift operators to construct a learnable graph topology and define a learnable graph Fourier basis by the singular value matrices using 1-D convolution (Conv-1D) neural layer. This eliminates the need for matrix inversion, thereby avoiding the associated numerical errors and stability problem.

eess.AS↗

arXiv2025

SEA-Spoof: Bridging The Gap in Multilingual Audio Deepfake Detection for South-East Asian

The rapid growth of the digital economy in South-East Asia (SEA) has amplified the risks of audio deepfakes, yet current datasets cover SEA languages only sparsely, leaving models poorly equipped to handle this critical region. This omission is critical: detection models trained on high-resource languages collapse when applied to SEA, due to mismatches in synthesis quality, language-specific characteristics, and data scarcity. To close this gap, we present SEA-Spoof, the first large-scale Audio Deepfake Detection (ADD) dataset especially for SEA languages. SEA-Spoof spans 300+ hours of paired real and spoof speech across Tamil, Hindi, Thai, Indonesian, Malay, and Vietnamese. Spoof samples are generated from a diverse mix of state-of-the-art open-source and commercial systems, capturing wide variability in style and fidelity. Benchmarking state-of-the-art detection models reveals severe cross-lingual degradation, but fine-tuning on SEA-Spoof dramatically restores performance across languages and synthesis sources. These results highlight the urgent need for SEA-focused research and establish SEA-Spoof as a foundation for developing robust, cross-lingual, and fraud-resilient detection systems.

cs.SD↗

arXiv2025

Long-Context Modeling Networks for Monaural Speech Enhancement: A Comparative Study

Advanced long-context modeling backbone networks, such as Transformer, Conformer, and Mamba, have demonstrated state-of-the-art performance in speech enhancement. However, a systematic and comprehensive comparative study of these backbones within a unified speech enhancement framework remains lacking. In addition, xLSTM, a more recent and efficient variant of LSTM, has shown promising results in language modeling and as a general-purpose vision backbone. In this paper, we investigate the capability of xLSTM in speech enhancement, and conduct a comprehensive comparison and analysis of the Transformer, Conformer, Mamba, and xLSTM backbones within a unified framework, considering both causal and noncausal configurations. Overall, xLSTM and Mamba achieve better performance than Transformer and Conformer. Mamba demonstrates significantly superior training and inference efficiency, particularly for long speech inputs, whereas xLSTM suffers from the slowest processing speed.

eess.AS↗

arXiv2025

Impact of Frame Rates on Speech Tokenizer: A Case Study on Mandarin and English

The speech tokenizer plays a crucial role in recent speech tasks, generally serving as a bridge between speech signals and language models. While low-frame-rate codecs are widely employed as speech tokenizers, the impact of frame rates on speech tokens remains underexplored. In this study, we investigate how varying frame rates affect speech tokenization by examining Mandarin and English, two typologically distinct languages. We encode speech at different frame rates and evaluate the resulting semantic tokens in the speech recognition task. Our findings reveal that frame rate variations influence speech tokenization differently for each language, highlighting the interplay between frame rates, phonetic density, and language-specific acoustic features. The results provide insights into optimizing frame rate selection for speech tokenizers, with implications for automatic speech recognition, text-to-speech, and other speech-related applications.

cs.CL↗

arXiv2025

Exploring Length Generalization For Transformer-based Speech Enhancement

Transformer network architecture has proven effective in speech enhancement. However, as its core module, self-attention suffers from quadratic complexity, making it infeasible for training on long speech utterances. In practical scenarios, speech enhancement models are often required to perform on noisy speech at run-time that is substantially longer than the training utterances. It remains a challenge how a Transformer-based speech enhancement model can generalize to long speech utterances. In this paper, extensive empirical studies are conducted to explore the model's length generalization ability. In particular, we conduct speech enhancement experiments on four training objectives and evaluate with five metrics. Our studies establish that positional encoding is an effective instrument to dampen the effect of utterance length on speech enhancement. We first explore several existing positional encoding methods, and the results show that relative positional encoding methods exhibit a better length generalization property than absolute positional encoding methods. Additionally, we also explore a simpler and more effective positional encoding scheme, i.e. LearnLin, that uses only one trainable parameter for each attention head to scale the real relative position between time frames, which learns the different preferences on short- or long-term dependencies of these heads. The results demonstrate that our proposal exhibits excellent length generalization ability with comparable or superior performance than other state-of-the-art positional encoding strategies.

eess.AS↗

arXiv2025

SpeechT-RAG: Reliable Depression Detection in LLMs with Retrieval-Augmented Generation Using Speech Timing Information

Large Language Models (LLMs) have been increasingly adopted for health-related tasks, yet their performance in depression detection remains limited when relying solely on text input. While Retrieval-Augmented Generation (RAG) typically enhances LLM capabilities, our experiments indicate that traditional text-based RAG systems struggle to significantly improve depression detection accuracy. This challenge stems partly from the rich depression-relevant information encoded in acoustic speech patterns information that current text-only approaches fail to capture effectively. To address this limitation, we conduct a systematic analysis of temporal speech patterns, comparing healthy individuals with those experiencing depression. Based on our findings, we introduce Speech Timing-based Retrieval-Augmented Generation, SpeechT-RAG, a novel system that leverages speech timing features for both accurate depression detection and reliable confidence estimation. This integrated approach not only outperforms traditional text-based RAG systems in detection accuracy but also enhances uncertainty quantification through a confidence scoring mechanism that naturally extends from the same temporal features. Our unified framework achieves comparable results to fine-tuned LLMs without additional training while simultaneously addressing the fundamental requirements for both accuracy and trustworthiness in mental health assessment.

eess.AS↗

arXiv2025

Mamba in Speech: Towards an Alternative to Self-Attention

Transformer and its derivatives have achieved success in diverse tasks across computer vision, natural language processing, and speech processing. To reduce the complexity of computations within the multi-head self-attention mechanism in Transformer, Selective State Space Models (i.e., Mamba) were proposed as an alternative. Mamba exhibited its effectiveness in natural language processing and computer vision tasks, but its superiority has rarely been investigated in speech signal processing. This paper explores solutions for applying Mamba to speech processing by discussing two typical speech processing tasks: speech recognition, which requires semantic and sequential information, and speech enhancement, which focuses primarily on sequential patterns. The experimental results confirm that bidirectional Mamba (BiMamba) consistently outperforms vanilla Mamba, highlighting the advantages of a bidirectional design for speech processing. Moreover, experiments demonstrate the effectiveness of BiMamba as an alternative to the self-attention module in the Transformer model and its derivates, particularly for the semantic-aware task. The crucial technologies for transferring Mamba to speech are then summarized in ablation studies and the discussion section, offering insights for extending this research to a broader scope of tasks.

eess.AS↗

arXiv2025

SAV-SE: Scene-aware Audio-Visual Speech Enhancement with Selective State Space Model

Speech enhancement plays an essential role in various applications, and the integration of visual information has been demonstrated to bring substantial advantages. However, the majority of current research concentrates on the examination of facial and lip movements, which can be compromised or entirely inaccessible in scenarios where occlusions occur or when the camera view is distant. Whereas contextual visual cues from the surrounding environment have been overlooked: for example, when we see a dog bark, our brain has the innate ability to discern and filter out the barking noise. To this end, in this paper, we introduce a novel task, i.e. SAV-SE. To our best knowledge, this is the first proposal to use rich contextual information from synchronized video as auxiliary cues to indicate the type of noise, which eventually improves the speech enhancement performance. Specifically, we propose the VC-S$^2$E method, which incorporates the Conformer and Mamba modules for their complementary strengths. Extensive experiments are conducted on public MUSIC, AVSpeech and AudioSet datasets, where the results demonstrate the superiority of VC-S$^2$E over other competitive methods. We will make the source code publicly available. Project demo page: https://AVSEPage.github.io/

cs.SD↗

arXiv2025

Should Audio Front-ends be Adaptive? Comparing Learnable and Adaptive Front-ends

Hand-crafted features, such as Mel-filterbanks, have traditionally been the choice for many audio processing applications. Recently, there has been a growing interest in learnable front-ends that extract representations directly from the raw audio waveform. \textcolor{black}{However, both hand-crafted filterbanks and current learnable front-ends lead to fixed computation graphs at inference time, failing to dynamically adapt to varying acoustic environments, a key feature of human auditory systems.} To this end, we explore the question of whether audio front-ends should be adaptive by comparing the Ada-FE front-end (a recently developed adaptive front-end that employs a neural adaptive feedback controller to dynamically adjust the Q-factors of its spectral decomposition filters) to established learnable front-ends. Specifically, we systematically investigate learnable front-ends and Ada-FE across two commonly used back-end backbones and a wide range of audio benchmarks including speech, sound event, and music. The comprehensive results show that our Ada-FE outperforms advanced learnable front-ends, and more importantly, it exhibits impressive stability or robustness on test samples over various training epochs.

eess.AS↗

arXiv2024

Time-Graph Frequency Representation with Singular Value Decomposition for Neural Speech Enhancement

Time-frequency (T-F) domain methods for monaural speech enhancement have benefited from the success of deep learning. Recently, focus has been put on designing two-stream network models to predict amplitude mask and phase separately, or, coupling the amplitude and phase into Cartesian coordinates and constructing real and imaginary pairs. However, most methods suffer from the alignment modeling of amplitude and phase (real and imaginary pairs) in a two-stream network framework, which inevitably incurs performance restrictions. In this paper, we introduce a graph Fourier transform defined with the singular value decomposition (GFT-SVD), resulting in real-valued time-graph representation for neural speech enhancement. This real-valued representation-based GFT-SVD provides an ability to align the modeling of amplitude and phase, leading to avoiding recovering the target speech phase information. Our findings demonstrate the effects of real-valued time-graph representation based on GFT-SVD for neutral speech enhancement. The extensive speech enhancement experiments establish that the combination of GFT-SVD and DNN outperforms the combination of GFT with the eigenvector decomposition (GFT-EVD) and magnitude estimation UNet, and outperforms the short-time Fourier transform (STFT) and DNN, regarding objective intelligibility and perceptual quality. We release our source code at: https://github.com/Wangfighting0015/GFT\_project.

eess.AS↗