Peer Revenue Publish or profit

Papers

Fresh from arXiv, typeset properly. Highlighted words trade on the market.

Electrical Engineering and Systems Science Audio and Speech Processing

eess.AS Watch Audio and Speech Processing

Tuesday 8 September

Audio and Speech Processing

Semantic Refinement of Universal Audio Representations through Audio-Description Alignment

Lejun Min, Junyu Dai, Ruichen Zheng et al. (+7 more)

Universal audio representations must preserve acoustic detail while making high-level concepts accessible across speech, music, environmental sound, and downstream models of different capacities. We study semantic refinement of an acoustically pretrained encoder by adding audio-description alignment to a foundation of BEST-RQ, reconstruction, and CTC. We compare matched control, shuffled-description, and correctly paired trajectories to distinguish correct correspondence from an extra contrastive objective. Each endpoint is frozen and evaluated with a temporal-mean linear probe and a sequence-aware LLM readout, testing whether the refined information is directly accessible and remains useful to a stronger model. Across three paired seeds, correct alignment improves domain-balanced classification by 4.66 points with the linear probe and 2.59 points with the sequence-aware LLM, with positive changes in every domain. Correct pairing accounts for 87% of the linear-probe gain, while the LLM shows its clearest correspondence-specific benefit in captioning. Dense acoustic objectives provide complementary gains under both readouts. A separate 24-layer continuation remains competitive with leading public encoders under the shared evaluator, supporting the recipe beyond the controlled study.
Read the full abstract

All authors Lejun Min, Junyu Dai, Ruichen Zheng, Xinyue Fan, Yang Xiang, Huaichen Zhang, Xingchen Song, Yufei Shi, Han Zhao, Xiangang Li

Audio and Speech Processing

Beyond Localisation Accuracy: Sensorimotor Effects of HRTF Individualisation

Fulvio Missoni, Katarina C. Poole, Tim Murray-Browne et al. (+2 more)

Everyday listening requires the brain to integrate cues from the body, environment, other senses, and movement, continuously translating auditory information into action. Yet HRTF individualisation is still commonly assessed through localisation accuracy, which may not fully capture its effects on this sensorimotor process. Here, we investigate whether these effects can instead be revealed through behaviour in a more ecologically valid listening task. We used an aurally guided visual search paradigm in which listeners located a visual target using a co-located virtual sound while moving freely, comparing individualised and non-individualised HRTFs under anechoic and reverberant conditions. Performance was assessed through response times and measures of movement organisation. In anechoic conditions, individualised HRTFs produced faster responses than non-individualised HRTFs, with an average reduction of approximately 200ms and the clearest benefit for front-back source locations. This advantage was expressed primarily in movement initiation, whereas overall movement extent was only weakly affected. Under reverberant conditions, HRTF-dependent differences disappeared. These results suggest that HRTF individualisation can influence how listeners plan and initiate orienting actions even when differences in conventional localisation outcomes are limited. Assessing sensorimotor behaviour alongside localisation performance may therefore provide a more sensitive and ecologically relevant account of the perceptual benefits of HRTF individualisation.
Read the full abstract

All authors Fulvio Missoni, Katarina C. Poole, Tim Murray-Browne, Andrea Canessa, Lorenzo Picinali

Audio and Speech Processing · Cross-listed from cs.CL

ConversationalVoice: Full-Duplex Speech Data from Real Conversations through Source-Faithful Reconstruction and Conversation-Grounded Expansion

Richard Yucheng He, Baodong Cao, Chen Xu et al. (+2 more)

Full-duplex speech models require training data that preserves turn-taking, overlap, interruption, and backchannel behavior, yet these signals are entangled across speakers in noisy real-world recordings. We present Conversational Voice, a pipeline that converts real two-speaker excerpts into three complementary training-data artifacts. (1) Separation recovers speaker-specific tracks with stable speaker assignments, a canonical transcript, and naturally observed interaction timing. (2) Reconstruction generates speech in matched voices from a fixed source transcript, reconstructs the source turn order, pauses, and overlaps, and adds word-level alignment and delivery instructions. (3) Expansion generates new dialogue constrained by the source context, speakers, and observed interaction pattern. Automatic speaker-verification metrics remain strong across stages, with same-speaker similarity of 0.983-0.991 and positive discrimination margins of 0.199-0.209. Predicted speech quality (NISQA MOS) is 3.56 for separation, 4.41 for reconstruction, and 4.61 for expansion. A Gemini-based automatic evaluator assigns expansion mean scores of 4.94/5 for contextual coherence and 4.80/5 for dialogue naturalness. Expansion and reconstruction exhibit broadly similar interaction profiles; expansion's turn, overlap-event, backchannel, and interruption rates are 4.6%, 8.0%, 13.2%, and 16.0% lower, respectively. We evaluate data properties only; downstream gains in full-duplex model training remain for future work.
Read the full abstract

All authors Richard Yucheng He, Baodong Cao, Chen Xu, Yihang Liu, Tairan Chen

Monday 7 September

Audio and Speech Processing · Cross-listed from cs.MM

Hi-Singers: A Comprehensive High-Quality Dataset for Expressive Audio-Driven Singing Head Synthesis

Yichi Zhang, Hui Zhang, Guanjun Liu et al. (+3 more)

State-of-the-art models for audio-driven digital human generation have achieved photo-realistic results in talking-head synthesis. However, extending these models to singing-head synthesis remains challenging due to a significant Domain Gap: singing requires more exaggerated expressions, vivid jaw openings, and precise rhythmic synchronization. Current models, primarily trained on speech datasets, often struggle with "rhythmic drift" and constrained dynamics. To address this, we introduce Hi-Singers, the first large-scale, high-quality, in-the-wild video dataset specifically tailored for singing head synthesis. Hi-Singers undergoes a rigorous automated and manual filtering pipeline, ensuring strict thematic adherence, high-resolution rendering, and stable motion, resulting in 29,608 video segments totaling approximately 170 hours. We further establish a dedicated evaluation benchmark balanced across linguistic and musical styles. Extensive experiments across diverse architectures, including 3D-coefficient and diffusion-based models, demonstrate that Hi-Singers consistently and significantly improves performance across all dimensions. Specifically, it enables models to achieve superior visual realism, enhanced lip-sync consistency, and more precise rhythmic dynamics, effectively bridging the domain gap and setting a new performance standard for the singing synthesis task. The dataset is available at https://huggingface.co/datasets/CharlesZhang-USTC/Hi-Singers
Read the full abstract

All authors Yichi Zhang, Hui Zhang, Guanjun Liu, Yuefeng Zou, Fengzhao Sun, Jun Yu

Audio and Speech Processing

Open-Set Vessel Re-Identification from Underwater Ship-Radiated Noise with a Raw-Waveform Selective-Kernel Acoustic Neural Network (SKANN) and a Cross-Passage Evaluation Protocol

Sunil Tyagi

Underwater acoustic target recognition has converged on closed-set classification by vessel type, a task that does not answer whether a monitoring system has heard this hull before. We formalise open-set, cross-passage vessel re-identification on public hydrophone data and specify a protocol that removes the two easiest routes to a high score: hull-disjoint splits keyed to MMSI/IMO, galleries and queries from disjoint passages of each hull, source-pure galleries, and an audio-adjudicated transit-deduplication gate. We describe SKANN, a raw-waveform encoder whose front end is a four-scale bank of learned filters fused by selective-kernel attention, trained with an angular-margin objective and an augmentation regime that perturbs recording chain, ambient noise and multipath while preserving the narrowband lines that carry identity. On a 40-hull IARA gallery (96 queries, 98 passage candidates), cross-passage rank-1 is 0.25 for the embedding and 0.26 for an automated narrowband-tonal comparator; the two are statistically indistinguishable at the top of the ranking, the embedding orders the rest of the list more reliably (AUC 0.82 vs 0.76), and their score fusion reaches rank-1 0.35 – the only contrast that attains nominal significance, presented as evidence of partial complementarity, not as a recommendation. Transit deduplication alone removes a 16-21 point apparent rank-1 advantage, larger than any between-method difference. Two further findings delimit what public data can support: ShipsEar cannot separate hull identity from recording channel under an identity protocol, and cross-network fine-tuning helps vessels seen during fine-tuning but is a null result on unseen ones. The results support analyst triage over a ranked shortlist, not identification. Checkpoint, validation embeddings, transit map and per-query outputs are released under CC-BY-4.0 (doi:10.5281/zenodo.22160138).
Read the full abstract

Audio and Speech Processing

Direction-Preserving Active Noise Control with a Conditional Control-Filter Estimation Network

Ziyi Yang, Zhengding Luo, Boxiang Wang et al. (+2 more)

Conventional active noise control (ANC) minimizes the total disturbance at the error microphone without distinguishing desired sound from noise. Direction-preserving ANC (DP-ANC) instead aims to attenuate a noise component arriving from a direction other than the specified desired direction while preserving sound naturally arriving from that direction. Existing approaches typically either require analytical optimization to be repeated for each new observation or estimate and reproduce the desired component through a hear-through secondary-source path. To address these limitations, this paper formulates DP-ANC as a direction-conditioned cancellation-preservation optimization problem. A component-separated objective jointly penalizes residual noise energy and the control response induced by the desired component, with a scalar weighting parameter controlling the cancellation-preservation trade-off. A convolutional network conditioned on the specified desired direction through feature-wise linear modulation (FiLM) is trained using a differentiable secondary-path-aware forward model. At deployment, the network estimates the complete multichannel finite impulse response (FIR) control-filter bank directly from a mixed-reference observation and the specified desired direction in a single forward pass, while retaining the conventional feedforward ANC signal path. Over 3300 evaluation cases, the selected operating point achieves 22.8 dB mean noise reduction with a desired-signal distortion of -11.4 dB. Validation using measured in-ear-device transfer functions further demonstrates consistent performance under measured acoustic configurations.
Read the full abstract

All authors Ziyi Yang, Zhengding Luo, Boxiang Wang, Libin Zhang, Woon-Seng Gan

Sunday 6 September

Audio and Speech Processing · Cross-listed from cs.MM

Causal Localization of the Refusal Direction in Audio Language Models

Leonardo Haw-Yang Foo and Hung-yi Lee

A large audio language model (LALM) attaches a speech front end to a text language model (LM) that is already safety-aligned. When such a model refuses a harmful spoken request, is the refusal carried by the front end, or inherited from the text LM? We test this with causal interventions. At each model's audio-to-LM interface and at tested LM residual layers, we fit a direction separating harmful from benign prompts, ablate its component, and measure the resulting change in the model's first-token refusal margin. Four of the five models are evaluated under held-out category shift. Across five LALMs spanning three backbone families, with three models passing a baseline safety gate, the largest tested effects occur in a mid-to-late LM band, while ablations of the tested interface directions have little effect. On Qwen2.5-Omni, ablating the L16 direction changes the margin by -7.10, versus -0.013 at the projector. The audio pathway is still in use: zeroing the encoder output changes the margin by -4.7. In the same model, the contrast is linearly decodable at an early layer where single-layer ablation has little effect, and a direction fitted on the LM backbone alone transfers to the full audio model. Because harmful and benign prompts also differ in form, we interpret the direction as refusal-linked rather than harmfulness-specific. These interventions localize dependence of the refusal margin, not where refusal is computed. Moving the margin also does not always change what the model writes. Safety audits of these models should use interventions rather than rely on probes alone, and should examine the inherited text LM alongside the audio interface.
Read the full abstract

Audio and Speech Processing · Cross-listed from cs.SD

From Masking to Merging: Rethinking SpecAugment for Efficient Audio Spectrogram Transformer

Minhee Park, Hyowon Ahn and Chanwoo Kim

This paper proposes SpecAugment-Patch Merging, a simple yet effective method to accelerate Audio Spectrogram Transformer (AST) training. We first apply SpecAugment to mask input spectrograms at the patch level, and after positional embeddings are added, the method selects r pairs of masked patches and merges them, reducing the number of tokens processed by the Transformer. Increasing the number of merged pairs r from 0 to 100 keeps mAP on AudioSet nearly unchanged (34.07 to 34.08) while throughput increases from 43.3 to 49.3 samples/sec, which is a relatively 13.9% improvement. Similar patterns appear on ESC-50 and Speech Commands V2, where throughput steadily improves with only minor accuracy changes, demonstrating that this merging approach provides faster training with minimal performance loss.
Read the full abstract

Audio and Speech Processing · Cross-listed from cs.SD

Lead Vocal Separation from Vocal Ensemble Mixtures Using Phoneme Alignment

Yuma Narahata, Tomohiko Nakamura, Yuki Saito et al. (+1 more)

Contemporary a cappella singing often has a lead-and-accompaniment texture, where the lead vocal (Vo) part carries the main melody and the remaining vocal parts provide accompaniment. Owing to their distinct roles, separating the Vo part from the remaining vocal parts, referred to as Vo separation, enables downstream applications such as lyric recognition and minus-one accompaniment generation for vocal ensemble music. Despite these potential applications, acoustic cues for this task are limited because the target and interfering sources are all singing voices with similar acoustic characteristics and often overlap in time, making Vo separation challenging. In this paper, we propose a Vo separation model that uses phoneme alignment of the Vo part as auxiliary information. The proposed model is based on band-split RoPE Transformer (BS-RoFormer), a state-of-the-art music source separation model, and introduces frame-level phoneme labels into its intermediate representations using feature-wise linear modulation (FiLM). Experimental results show that phoneme-alignment conditioning improves Vo separation performance over an audio-only baseline and yields larger average gains than conditioning only on Vo singing/silence activity. Further analysis suggests that the advantage of phoneme-label information is larger when fewer remaining vocal parts share the same phoneme as Vo.
Read the full abstract

All authors Yuma Narahata, Tomohiko Nakamura, Yuki Saito, Hiroshi Saruwatari

Saturday 5 September

Audio and Speech Processing

SETEAB: Multiscale approach with Squeeze-and-Excitation Temporal Enhanced Aware Block for Speech Emotion Recognition

Duy Vo, Kiet Anh Hoang and Hao Do

This paper proposes a novel lightweight multiscale architecture for speech emotion recognition (SER) with three key innovations. First, a depthwise convolution-based subsampling module is introduced to reduce model size and computation while preserving salient emotional cues. Second, a Squeeze-and-Excitation block is integrated to enhance channel-wise recalibration and improve representation robustness. Third, a new Temporal Enhanced Aware Block is designed to strengthen temporal dependency modeling and produce more discriminative emotion-aware features. The proposed model is explicitly designed to jointly improve compactness, recognition performance, and generalizability. Experiments on benchmark SER datasets show that our method achieves higher accuracy with reduced computational complexity, while also delivering stronger cross-corpus performance than most recent advanced networks for SER.
Read the full abstract

Friday 4 September

Audio and Speech Processing

Enhancing Neural Speech Coding with Semantic and Visual Cues

Yao Guo, Yang Ai, Hui-Peng Du et al. (+3 more)

At low bitrates, neural speech codecs have limited capacity to encode all information needed for high-quality re construction, especially when relying solely on speech-derived representations. To address this limitation, this paper proposes a Semantic- and Visual-enhanced Speech Codec (SVSC), which in corporates semantic and visual cues into the neural speech coding process. Specifically, built upon a mainstream neural speech cod ing architecture, SVSC introduces a semantic encoding-decoding branch and an image analysis-synthesis branch. It fuses deep semantic features with visual cues through a cross-attention mech anism, forming an auxiliary high-level representation enriched with contextual and articulatory information. To handle different inference scenarios, SVSC introduces two information-injection strategies based on the availability of auxiliary semantic and vi sual cues. When such cues are available, the fusion mode directly incorporates the auxiliary representations into the speech coding branch through feature concatenation; otherwise, the distillation mode transfers auxiliary information into the speech coding branch through knowledge distillation during training, enabling speech-only inference without additional inputs. Experimental results validate the effectiveness of incorporating semantic and visual cues, improving the ViSQOL score of reconstructed speech from 3.86 to 4.01.
Read the full abstract

All authors Yao Guo, Yang Ai, Hui-Peng Du, Xiao-Hang Jiang, Chen-Yuan Ning, Zhen-Hua Ling

Audio and Speech Processing · Cross-listed from eess.SP

What Selects, What Reconstructs: Repairing Exemplar-Based Complex-Spectrum Separation

Maxime Baelde

Exemplar methods separate a mixture by picking one learned spectrum per source and deforming it until it explains the observation, making one deformation class both reconstructor and selector. We show that the second role is empty as soon as the class can interpolate: the rule then ranks candidates on its regulariser, a choice made before the data, and the estimates sum back to the mixture whichever candidate wins. The condition is a parameter count, so the diagnosis runs before any experiment. On free per-bin deformation of complex spectra it explains the observed pathologies at once: a criterion that ranks candidates by their loudness, and half an output that is a mask on the mixture rather than an exemplar. The same theorem prescribes the repair, a selection class poorer than the reconstruction class: one complex gain and one pure delay rank the candidates, and a local combination of the best-aligned atoms, fitted jointly in closed form, rebuilds them. On MUSDB18 against the exact ceiling of the masking class, the distance between the criterion and an oracle inside its own candidate pool falls under the rigid selector from 6.2-7.7 to 0.5-2.8 dB, though only 0.9-1.2 dB of that reaches the output, and the per-frame latency of the deployed rule by a factor of 47 to 806. One lock remains, quantified: atoms are scored against the mixture, so the score carries a term for the other source that absorbs the capacity the reconstruction class gains, leaving the output 10.0 dB under the ceiling. Ranking hypotheses by the residual of a fit free enough to interpolate ranks them on the regulariser alone.
Read the full abstract

Thursday 3 September

Audio and Speech Processing

Discriminative Flow Matching: Beyond Time-Conditioning in Generative Restoration via Flow-State Representations

Shrishti Saha Shetu, Emanuël A. P. Habets and Andreas Brendel

Existing Conditional Flow Matching (CFM) formulations describe transport progress using an explicit interpolation coordinate, commonly interpreted as time, assuming that a single global variable adequately represents a sample's position along the generative trajectory. In restoration tasks, however, transport progress is sample-dependent because the initial distribution may exhibit varying statistical dependencies with the target distribution. Thus, samples at the same interpolation coordinate can differ substantially in degradation level, distance to the target distribution, and restoration difficulty. We investigate whether signal representations learned by discriminatively trained models provide a meaningful description of generative transport state in CFM-based restoration. Through systematic latent-space analysis, we show that discriminative representations organize according to degradation severity and follow a consistent trajectory toward the clean-data manifold during generation. Motivated by these observations, we introduce the Discriminative Flow-State Hypothesis, which posits that discriminative representations encode a transport state governing generative restoration. Based on this hypothesis, we propose Discriminative Flow Matching, which conditions the Flow-Matching velocity field on Discriminative Flow-State Representations rather than explicit time coordinates. Experiments on speech enhancement and image denoising show that these representations characterize restoration progress, enable adaptive inference, and consistently outperform CFM and diffusion-related baselines. Our findings suggest that discriminative representations provide an effective state-aware alternative to explicit time conditioning and offer a novel perspective on the relationship between discriminative and CFM-based generative modeling.
Read the full abstract

Audio and Speech Processing

Brain2Speech-Net: Fast and Intelligible Brain-to-Speech Synthesis Without Text Decoding

Shreeram Suresh Chandra, Zexin Cai, Yu Tsao et al. (+2 more)

The loss of speech limits communication for individuals with paralysis. Direct neural-to-speech synthesis is challenging due to the limited availability of neural data for training speech brain-computer interfaces. Most existing systems rely on cascaded neural-to-text-to-speech pipelines, which increase inference latency and propagate errors across stages. We present Brain2Speech-Net, a single-stage neural-to-speech generation framework without intermediate text decoding. We use a differentiable phoneme bottleneck and a deep-HMM alignment mechanism to map long neural recordings into the latent space of a text-to-speech (TTS) model, enabling high-quality speech synthesis. Brain2Speech-Net is the only system in our comparison that produces intelligible speech while generating faster than real time.
Read the full abstract

All authors Shreeram Suresh Chandra, Zexin Cai, Yu Tsao, Simon King, Berrak Sisman

Audio and Speech Processing · Cross-listed from cs.SD

Scalable Context Orchestration for Serving LLMs Over Voice

Linyi Jiang, Silvery D. Fu and Yifei Zhu

Voice AI applications are gaining popularity as advances in large language models (LLMs) enable more natural and accessible spoken interactions. Serving these applications requires accounting not only for what users say, but also for how they speak (e.g., speaking rate) and the conditions under which their audio is captured and transmitted (e.g., background noise and packet loss). However, existing LLM systems represent conversation context as a flat, growing sequence of messages, leaving voice-specific context implicit in the audio. As a result, they can generate responses that are poorly aligned with user preferences, degrade interaction quality under adverse environmental conditions, and incur high costs over long voice sessions. We present llmovoice, a context-management middleware that explicitly models voice context and orchestrates its use. At each turn, llmovoice constructs a bounded voice context from the current user input, relevant interaction history, and explicit paralinguistic and environmental states. It then uses the serving LLM to reason over this context and generate runtime directives that guide how the system responds. We evaluate llmovoice on real-world voice applications and benchmarks. It reduces speaking-rate alignment error by 52.4%, lowers the false-interruption rate from 46.0% to 0.9% under packet loss, and reduces model usage cost by 79.2%. For long sessions, llmovoice reduces per-turn cost by up to 24.9 times while retaining up to 98.7% of baseline answer quality.
Read the full abstract

Audio and Speech Processing

Deep Neural Compression for RIR-Characterized Acoustic Environments with Structure-Aware Constraints

Chen-Yuan Ning, Yang Ai, Hui-Peng Du et al. (+2 more)

Room impulse responses (RIRs) characterize the acoustic environment of a room by capturing how sound propagates and decays within an enclosed space. In applications such as immersive audio rendering, accurate acoustic reconstruction often relies on spatially densely sampled RIRs. This consequently gives rise to a large volume of RIR data, imposing a substantial burden on storage. Although recent neural audio codecs provide an effective framework for low-bitrate compression, their training objectives are mainly tailored to speech and general audio, and are therefore not well aligned with the acoustic characteristics of RIRs. Therefore, we propose an EnCodec-based neural RIR compression method, which incorporates RIR structure-aware constraints at two levels. Specifically, at the RIR level, structure-aware constraints are imposed on the global decay behavior and local energy distribution of RIRs through energy decay curve (EDC) regularization and a short-time window energy constraint, while at the reverberant-speech level, reverberant-speech supervision is further introduced to constrain the consistency of the reverberant speech generated by the reconstructed RIRs. Experimental results show that, at a low bitrate of 375 bps, the proposed method achieves lower RIR reconstruction error and better reverberant-speech perceptual consistency than audio-oriented codecs.
Read the full abstract

All authors Chen-Yuan Ning, Yang Ai, Hui-Peng Du, Xiao-Hang Jiang, Zhen-Hua Ling

Audio and Speech Processing · Cross-listed from cs.CL

Alignment-Free Text-Audiobox for Voice Dubbing and Full-Duplex Dialogue Synthesis

Sanyuan Chen, Min-Jae Hwang, Sho Inoue et al. (+12 more)

We present Alignment-Free Text-Audiobox (Text-AB), a unified framework for high-quality voice dubbing and full-duplex dialogue synthesis. Building on a Diffusion Transformer trained with a flow-matching objective, Text-AB departs from the Audiobox system along three dimensions. First, it operates in a latent diffusion framework using DAC-VAE features that encode 48 kHz waveforms into a 25 Hz latent sequence, giving over 10x higher compression than previous EnCodec representations while improving resynthesis quality. Second, Text-AB is alignment-free: it consumes raw text via an off-the-shelf text encoder and learns text-speech alignment through cross-attention, removing the need for forced alignment and explicit duration prediction. Third, we scale model and data substantially, pretraining a 3B-parameter model on 480k hours of monolingual speech, followed by supervised fine-tuning on three downstream tasks: cross-lingual voice dubbing, full-duplex dialogue synthesis, and emotional full-duplex dialogue synthesis. At inference, Text-AB supports one-shot generation for up to ~1 min of speech and arbitrarily long-form generation via a multi-diffusion scheme, plus a multi-stage reranking strategy that enhances quality based on automated metrics. On a real-world dubbing benchmark, Text-AB delivers a step-change improvement over the latest internal dubbing system, with large gains in prosody similarity, voice similarity, naturalness, and shareability. For full-duplex dialogue synthesis, it approaches human recordings on short-form conversations and substantially outperforms the latest internal model on long-form human-likeness and expressivity, while natively modeling turn-taking, back-channeling, and emotional dynamics. For emotional dialogue synthesis, emotion conditioning significantly improves emotion alignment and emotional interaction quality over the unconditioned baseline.
Read the full abstract

All authors Sanyuan Chen, Min-Jae Hwang, Sho Inoue, Anna Sun, Bokai Yu, David Kant, Dongmin Hyun, Dorian Desblancs, Gregory Antonovsky, Oleg Repin, Peng-Jen Chen, Xutai Ma, Zehai Tu, Juan Pino, Wei-Ning Hsu

Audio and Speech Processing

Fairness Evaluation of Edge-AI Implementation for Cleft Lip and Palate Speech ASR

Susmita Bhattacharjee, Himashri Deka, H.S. Shekhawat et al. (+1 more)

Automatic speech recognition (ASR) remains challenging for individuals with cleft lip and palate (CLP) because of limited pathological speech data and large variations in speech characteristics across speakers and severity levels. These recognition difficulties can reduce the accessibility of voice-based human-computer interaction, particularly when cloud-based ASR services are unavailable or unreliable. This work investigates a severity-aware and edge-deployable ASR framework for improving recognition of CLP speech using Whisper-small. The model was fine-tuned using different combinations of normal and CLP speech representing mild, moderate, and severe conditions, together with a CLP-only training configuration, to examine how the inclusion of different severity levels influences recognition performance and fairness across speakers. The pretrained model produced pooled word error rate (WER) and phoneme error rate (PER) values of 62.46% and 52.72%, respectively. Severity-aware fine-tuning substantially improved performance, reducing the best pooled WER to 22.72% and the best pooled PER to 18.44%. Training with a broader representation of CLP severity levels also provided the best overall balance between recognition accuracy and performance consistency across severity groups. Deployment on an NVIDIA Jetson platform demonstrated real-time inference for all fine-tuned models, with real-time factors of 0.167-0.171 and peak GPU memory usage of approximately 566 MB. The results demonstrate that incorporating severity diversity during ASR adaptation can substantially improve recognition of CLP speech while reducing performance disparities across severity groups. The proposed approach further enables low-latency, Internet-independent speech interaction on edge devices, supporting more accessible and inclusive voice-based human-computer interaction for individuals with CLP.
Read the full abstract

All authors Susmita Bhattacharjee, Himashri Deka, H.S. Shekhawat, S.R.M. Prasanna

Audio and Speech Processing

ToolDF: Tool-Integrated Reasoning for Mixed-Authenticity Audio Deepfake Detection

Taewoo Kim, Young Han Lee, Nam In Park et al. (+1 more)

Audio deepfake detection is commonly formulated as clip-level binary classification of single-domain audio. However, real-world manipulated audio can exhibit mixed authenticity, where genuine and manipulated cues coexist across temporal transitions, overlapping sources, or both. This setting requires not only detecting manipulated audio but also localizing the components that provide evidence for the decision. We propose ToolDF, a tool-integrated reasoning framework for mixed-authenticity audio deepfake detection. ToolDF employs an audio large language model as an orchestrator trained with supervised tool-use trajectories. It adaptively analyzes the audio scene, selectively performs source separation, routes components to domain-specific experts, and aggregates their evidence into an interpretable verdict. We further introduce a mixed-authenticity ADD benchmark covering temporal transitions, acoustic overlaps, and hybrid mixtures. Experimental results show that ToolDF achieves the best overall performance on composite-type detection, achieving macro-F1 gains of 3.72 and 14.39 points over the strongest monolithic baseline and a fixed pipeline, respectively, while providing interpretable evidence localized to temporal regions and acoustic sources. Our source code and dataset are publicly available online.
Read the full abstract

All authors Taewoo Kim, Young Han Lee, Nam In Park, Chanwoo Kim

Audio and Speech Processing

Broadband Acoustic Intensity Direction Estimation with Tight-Frame Cardioid Arrays

Akira Omoto

This study experimentally investigates source-direction estimation from acoustic intensity using paired cardioid microphones, with particular emphasis on well-balanced arrangements known as tight frames. Impulse responses were measured at 6 or 24 microphone positions, and two opposed-pair spacings were examined. The resulting arrays were evaluated for a single source, coherent interference between waves arriving from two orthogonal directions, and directional tracking in the presence of interfering waves from multiple directions. The results demonstrate broadband acoustic-intensity direction estimation from 50 Hz to 20 kHz, with the 24-microphone arrangement and the shorter spacing generally providing smaller errors.
Read the full abstract

Audio and Speech Processing · Cross-listed from eess.SP

Geometric Ceilings on Time-Frequency Masking for Single-Channel Separation

Maxime Baelde

Most single-channel separators estimate a source by applying a real gain to the mixture in each time-frequency bin. The optimum of that format, which the oracle masks used as bounds do not attain, is the orthogonal projection of the source onto the line spanned by the mixture, its residual set by the angle between them. Locating an estimator reduces to the block structure of a real-linear operator on stacked spectra, giving a chain of four nested classes whose three larger terms match three assumptions on the prior: zero means, circularity and absence of inter-frequency coupling. Held fixed the chain is a cascade of four orthogonal projections; refitted per frame it collapses onto its first term, attributing the whole residual to one missing real parameter per bin, the phase. When the phase posterior is symmetric about the mixture direction, the minimum mean-square estimate falls back onto the line, with gain the posterior mean of the oracle gain and excess error its variance. On MUSDB18 a posterior mean under a non-circular Gaussian-mixture prior leaves the class yet stays 11.44 dB under the per-frame ceiling, which four times as many components and 7.5x the data do not close; a closed-form gate attributes some 70% of it, in decibels, to the predicted variance. The widest fixed class stays 6.70 dB under the same ceiling. Leaving the class and minimising squared error are conflicting requests: the barrier lies in the criterion rather than in the prior.
Read the full abstract

Audio and Speech Processing

Summary of the ChinaVoices Challenge 2026: Data, Tasks, Baseline, and Methods

Yujie Liao, Bingshen Mu, Shuiyuan Wang et al. (+5 more)

This paper summarizes the ChinaVoices Challenge 2026, which aims to establish unified task definitions and evaluation conditions for Chinese dialect speech processing and to advance multi-dialect identification and automatic speech recognition. The challenge covers 16 dialect categories and defines two tasks: Chinese Multi-Dialect Identification and Chinese Multi-Dialect Automatic Speech Recognition (ASR). It uses approximately 320 hours of speech across the Reference Set, Open Evaluation Set, and Hidden Evaluation Set. The two tasks use the same evaluation audio, and each includes restricted-data and open-data tracks. We describe the task settings, data, evaluation metrics, and Qwen3-ASR-1.7B baseline, and analyze the leaderboard results and submitted systems. In total, 28 teams submit results, 17 provide system reports, and systems from 15 teams pass the compliance review and are included in the analysis. Most eligible systems outperform the baseline, and the official top-three order remains unchanged on the Hidden Evaluation Set for both tasks. Dialect-level results show that categories with higher identification accuracy generally have lower ASR error rates, although the tasks assess related but distinct capabilities. Leading identification systems commonly exploit dialect-discriminative acoustic representations, whereas leading ASR systems emphasize data normalization, augmentation, and auxiliary CTC objectives. These results provide practical guidance for developing and evaluating Chinese multi-dialect speech processing systems.
Read the full abstract

All authors Yujie Liao, Bingshen Mu, Shuiyuan Wang, Liumeng Xue, Hexin Liu, Xian Shi, Jie Hu, Lei Xie

Audio and Speech Processing

StreamWSR: Streamable and Lightweight Waveform-Domain Neural Speech Super-Resolution

Yuan Tian, Yang Ai, Hui-Peng Du et al. (+1 more)

This paper proposes StreamWSR, a Streamable neural Waveform-domain model for speech Super-Resolution (SR). By adopting a fully causal architecture with compact frame-level waveform representation, the proposed StreamWSR supports zero-look-ahead streaming inference while avoiding vocoder-based reconstruction and explicit phase prediction. Specifically, StreamWSR downsamples the input waveform into a compact frame-level representation using strided causal convolutions. Then, a lightweight causal long-short-term modeling backbone is employed to capture both local waveform structures and long-range historical dependencies under causal constraints. Finally, the modeled output is converted back to the waveform domain through a causal transposed-convolution and combined with the input waveform via a residual connection to generate the final high-resolution speech. Experimental results on 16 kHz speech SR show that StreamWSR achieves competitive or superior speech quality and intelligibility compared with representative waveform- and spectrum-based baselines, while maintaining a zero-look-ahead streaming advantage with only 9M parameters and 2G FLOPs.
Read the full abstract

All authors Yuan Tian, Yang Ai, Hui-Peng Du, Zhen-Hua Ling

Wednesday 2 September

Audio and Speech Processing · Cross-listed from cs.SD

Low-Latency Spell Correction for Japanese Music Search Queries

Anshul Garg, Pavni Tandon, Karan Bhukar et al. (+2 more)

Spell correction for Japanese search queries presents unique challenges due to the co-existence of four writing scripts (Latin/romaji, hiragana, katakana, and kanji) and the distinct error patterns each script induces. We present a compact BART-based sequence-to-sequence model (3 encoder + 3 decoder layers) designed for low-latency spell correction of Japanese music search queries. The core contribution lies in a script-aware synthetic misspelling generation pipeline that produces realistic training data by combining keyboard-layout models (QWERTY and flick input), phonetic confusion priors mined from real query logs, voiced/unvoiced consonant alternations, and kana case errors. A key design decision is normalizing mixed-script catalog titles to a single canonical script before misspelling synthesis, which we show is critical for reducing model hallucinations. We train a custom byte-level BPE tokenizer on the target music catalog to handle all four scripts in a unified vocabulary. Experiments on a curated evaluation set show that our model achieves an exact-match accuracy of 41.09% and a character error rate (CER) of 11.62%, outperforming edit-distance baselines and achieving the lowest character error rate among all evaluated systems while maintaining sub-4ms inference latency on a single GPU. We further analyze performance across individual scripts and mixed-script queries, demonstrating the effectiveness of script-aware data augmentation through systematic ablation studies.
Read the full abstract

All authors Anshul Garg, Pavni Tandon, Karan Bhukar, Tanmay Khandelwal, Ujjal Kumar Dutta

Audio and Speech Processing

GhostWord: A Fine-Grained Backdoor Attack on Automatic Speech Recognition

Mojtaba Nafez, Mobina Poulaei, Kiarash Kiani Feriz et al. (+4 more)

Automatic Speech Recognition (ASR) systems are widely deployed in safety-critical settings but remain vulnerable to data-poisoning backdoor attacks. Existing ASR backdoors typically use phrase-level triggers paired with a fixed target sentence, creating strong artifacts (e.g., repeated transcripts or triggers placed in non-speech regions) that simple preprocessing can mitigate. We propose GhostWord, a word-level, time-localized ASR backdoor that uses codebooks mapping short ($\approx$400\,ms) acoustic triggers to target words. During poisoning, we inject a trigger into the forced-aligned time span of a chosen source word in the audio and replace only that word in the transcript, enabling precise semantic flips and composable sentence manipulation while avoiding many-to-one label artifacts. Across Common Voice (v23 English, v24 Lithuanian) and multiple backbones (Whisper-Small/Medium, MMS, SpeechT5), GhostWord achieves an average attack success rate of 89.3% and transfers across languages and models. Adapting optimization-based defenses (ABL, ANP, SAU, I-BAU) reveals a sharp robustness–accuracy trade-off: attack success drops from 89.3% to 29.1% while clean WER rises from 21.5% to 45.0%, consistent with our theoretical analysis showing that, in high-vocabulary models, backdoor suppression structurally tends to degrade clean performance. The source code is publicly available at https://github.com/rohban-lab/GhostWord
Read the full abstract

All authors Mojtaba Nafez, Mobina Poulaei, Kiarash Kiani Feriz, Aref Mousavi, Mohammad Ebrahim Mahdavi, Mohammad Mosayebi, Mohammad Hossein Rohban

Audio and Speech Processing · Cross-listed from cs.SD

VoxReason: Auditing Source-Grounded Speech Plans Before Synthesis

Mengzhe Geng

Plan accuracy alone cannot show whether a speech-delivery decision follows its source: a fixed prior may match the original label yet fail to respond appropriately when a cue changes. VoxReason provides a 100-case verifier benchmark that holds each utterance fixed, edits one designated source-label cue, and scores cited evidence, eight plan fields, and the permitted response. On a source-key-disjoint test of 24 cases, a source-emotion prior reaches plan-slot accuracy 0.958, but none of the 24 edited neutral targets appears in its training labels; its required-change accuracy is 0.000. This diagnoses the support boundary of this prior, not its performance on supported edits. In a complementary 32-case emotion-disjoint test, the prior has seen all edited neutral targets but neither original test emotion; its plan-slot accuracy is 0.219 and required-change accuracy is 1.000. The partitions reuse and overlap the same 100 cases, so these deterministic diagnostics are not independent cohorts or learned-planner results. The benchmark evaluates derived labels and structured plans, not audio input, generated speech, or listener judgments.
Read the full abstract

Audio and Speech Processing

VibeVoice-ASR-Streaming Technical Report

Yujie Tu, Zhiliang Peng, Jianwei Yu et al. (+11 more)

Traditional speaker-attributed ASR systems treated ASR and speaker diarization as two separate tasks. Recently, end-to-end models such as VibeVoice-ASR have unified the two tasks within a single model. However, existing unified models still mainly support offline recognition, making it difficult to meet the low-latency requirements of real-time voice assistants and agents. To tackle this issue, we present VibeVoice-ASR-Streaming, one of the first LLM-based end-to-end approaches to streaming speaker-attributed ASR. It interleaves fixed-size audio chunks, a small amount of lookahead audio and previous text. This allows the model to produce ”who said what” as speech arrives, without a separate diarization stage. For transcription accuracy, our 7B model achieves the lowest average WER/CER across five evaluation sets. For speaker attribution, it achieves the best or tied-best on 12 of 13 evaluation settings. We release the 1.5B and 7B model weights together with inference code.
Read the full abstract

All authors Yujie Tu, Zhiliang Peng, Jianwei Yu, Li Dong, Songchen Xu, Yaoyao Chang, Wenhui Wang, Zilong Wang, Zehua Wang, Yan Xia, Ruibin Yuan, Jiajun Zhang, Xie Chen, Furu Wei

Audio and Speech Processing · Cross-listed from eess.SP

ARFT: A Synchronized Multimodal RF-Acoustic Dataset for Positioning in Distributed Environments

Daan Delabie, Jarne Van Mulders, Bert Pyck et al. (+2 more)

This paper documents the acoustic-radio fusion in Techtile (ARFT) dataset, a synchronized measurement campaign for distributed wireless sensing and positioning in the Techtile testbed. Ultrasonic and radio frequency (RF) signals are simultaneously transmitted and captured at multiple positions in a 2D spatial grid inside the Techtile testbed. Each acquisition cycle corresponds to one rover stop, one position sample, one acoustic recording and one RF snapshot. The RF modality is recorded as per-host pilot measurements and released as channel state information (CSI) tensors for 42 antennas mounted at the ceiling of the room. The acoustic chirp is recorded with 91 synchronized microphones and transmitted by a synchronized quasi omni-directional speaker. The campaign spans 5011 spatial position samples spanning a 5.57 m by 2.89 m area with complete RF and acoustic data. We detail how the measurements are recorded, quantify per-experiment coverage, describe the RF and acoustic data contents, and explain how the aligned modalities support RF-only, acoustic-only, and joint positioning workflows. Furthermore, a static acoustic positioning pipeline that performs anchor selection, pulse-compression ranging, and least squares (LS) localization is elaborated.
Read the full abstract

All authors Daan Delabie, Jarne Van Mulders, Bert Pyck, Gustav Nilsson Gisleskog, Gilles Callebaut

Audio and Speech Processing

VAANI Noise Event Dataset: A curated spontaneous speech dataset annotated with timestamps for noise events

Pavan Kumar J, Agneedh Basu, Pranav Bhat et al. (+3 more)

Most public sound-event corpora are optimized either for general audio tagging or for clean speech separation, and comparatively few provide strong timestamped noise annotations layered directly on top of spontaneous, real-world speech. We present the VAANI Noise Event Timestamp Dataset, a derived annotation layer built on Project VAANI field recordings of spontaneous speech collected across 165 Indian districts in 105 languages. Unlike synthetically mixed corpora, VAANI captures speech and ambient noise in situ and simultaneously, and annotates each recording with fine-grained start/end timestamps for overlapping background noise events organized into a compact seven-class semantic taxonomy: animal, traffic, baby/child, music, signal/alarm, appliance, and non-speech human. This combination of spontaneous multilingual Indic speech, authentic regional soundscapes, and span-level noise tags that may overlap with speech targets tasks that existing datasets address only partially: noise-robust Automatic Speech Recognition (ASR), sound event detection (SED), and speech enhancement. We position VAANI against nine widely used corpora and benchmarks, including WHAM!, AVA-Speech, MUSAN, FSD50K, CHiME-6, AudioSet, DESED, the India-specific iNoise noise database, and the Kathbath-Noisy noisy-ASR benchmarks, and describe the annotation protocol and quality-control procedure used to produce the timestamped tags.
Read the full abstract

All authors Pavan Kumar J, Agneedh Basu, Pranav Bhat, Sujith Pulikodan, Suryansh Shukla, Nihar Desai Prasanta K. Ghosh

Audio and Speech Processing · Cross-listed from cs.SD

SonicCaps: Large-Scale Diverse and Fine-Grained Captioning for Improved Audio-Retrieval

Zineb Lahrichi, Marc Ferras, Gaël Richard et al. (+1 more)

Recent advances in audio-language modeling have been driven by large-scale audio captioning datasets. However, existing datasets remain limited by low semantic diversity, generic descriptions lacking acoustic details, and one-to-one audio-caption mappings that poorly reflect the inherent ambiguity of auditory perception. We introduce SonicCaps, a large-scale audio captioning dataset comprising ~15M captions paired with ~700k audio clips, generated using a multi-modal large language model (Qwen3-Omni) conditioned on both audio and text. To explicitly promote diversity, we generate around 24 captions per audio via structured prompt engineering and few- shot generation, spanning main descriptions, rephrased variants (verbosity, style) and semantic tags. Human evaluation shows that SonicCaps is rated significantly higher than existing captioning datasets, with fine-grained analyses indicating that our captions are perceived as more descriptive and precise, which strongly correlates with quality judgments. Finally, training CLAP models on SonicCaps with a multi-caption sampling strategy consistently improves audio retrieval and zero-shot classification, with stronger generalization across public and commercial benchmarks. We release both SonicCaps and two specialized CLAP models on hugging face: https://huggingface.co/datasets/Zineb/SonicCaps.
Read the full abstract

All authors Zineb Lahrichi, Marc Ferras, Gaël Richard, Geoffroy Peeters