Peer Revenue Fully funded

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

Wednesday 12 August

Audio and Speech Processing

On-Policy Self-Distillation for Multi-Dialect ASR: Mastering Dialects, Retaining Mandarin

Shuiyuan Wang, Bingshen Mu, Pengshen Zhang et al. (+6 more)

Recent large-scale ASR models already achieve strong Mandarin recognition accuracy and have some ability to recognize Chinese dialects. However, their dialect recognition accuracy is still limited in real-world speech. Direct dialect adaptation can lower dialect CER, but it may also raise Mandarin CER. We therefore study how to adapt a capable ASR model to improve multi-dialect recognition without degrading Mandarin recognition. We adopt an adaptation pipeline where continual pre-training (CPT) and dialect supervised fine-tuning (SFT) provide a strong foundation, and On-Policy Self-Distillation (OPSD) serves as the final refinement. OPSD addresses the train–test mismatch in autoregressive ASR by training the student model on its own decoded prefixes while a frozen teacher, conditioned on the reference transcript as privileged context, provides soft token-level targets. This replaces hard cross-entropy updates on dialect data with distillation, preserving Mandarin ability while refining dialect recognition. We instantiate the framework with Qwen3-ASR-1.7B and evaluate it on public and internal Mandarin and dialect test sets. Under matched refinement data and schedule, OPSD improves dialect recognition without raising Mandarin CER, whereas continued teacher-forced fine-tuning increases Mandarin CER. We will release the model weights and evaluation scripts.
Read the full abstract

All authors Shuiyuan Wang, Bingshen Mu, Pengshen Zhang, Chengyou Wang, Yujie Liao, Chengdong Liang, Binbin Zhang, Qiangze Feng, Lei Xie

Audio and Speech Processing

MiDashengLM-Gen: Unified Audio Scene Generation via LLM-Driven Autoregressive Flow Matching

Xingwei Sun, Heinrich Dinkel, Gang Li et al. (+7 more)

Generating coherent audio scenes that simultaneously blend speech, music, and sound effects remains a significant challenge. Current approaches typically rely on a disjointed pipeline where a frozen, decoupled text encoder feeds a separate audio decoder, limiting cross-modal optimization and leading to poor speech intelligibility. To overcome these limitations, we introduce MiDashengLM-Gen, an end-to-end framework that couples a pre-trained Large Language Model (LLM) with per-token conditional flow matching for autoregressive, variable-length mixed-audio scene generation. MiDashengLM-Gen represents a first approach for general text-to-audio generation with one end-to-end trained model. Empirical evaluations demonstrate that MiDashengLM-Gen drastically improves speech intelligibility over existing unified models. On the Seed-TTS benchmark, English Word Error Rate (WER) drops from 12.15% to 2.79%, approaching the performance of dedicated Text-to-Speech (TTS) systems (1.24%). Furthermore, the framework extends effectively to multilingual settings, yielding highly competitive multilingual WERs compared to existing baselines. Lastly, the model maintains competitive mixed-audio generation quality on the MECAT benchmark. Code and checkpoints are available at https://github.com/xiaomi-research/midashenglm-gen and https://huggingface.co/mispeech/midashenglm-gen, and the demo page is available at https://xingws.github.io/midashenglm-gen-demo/.
Read the full abstract

All authors Xingwei Sun, Heinrich Dinkel, Gang Li, Jiahao Mei, Yadong Niu, Zerui Han, Yuepeng Jiang, Jiahao Zhou, Lichun Fan, Jian Luan

Audio and Speech Processing

Deep Learning Based Relative Transfer Matrix Estimation for Multiple Sources and Multiple Microphones

Oshan A. B. Yalegama and Wageesha N. Manamperi

The Relative Transfer Matrix (ReTM), recently introduced as a generalization of the relative transfer function for multiple receivers and sources, shows promising performance when applied to speech enhancement in noisy environments. Estimating the ReTM of sound sources by exploiting the covariance matrices of multichannel recordings is highly beneficial for practical applications and, to date, remains the only proposed approach. This paper investigates deep learning-based ReTM estimation. We propose three novel supervised learning frameworks using time and short-time frequency transform domain convolutional networks, and a Long Short-Term Memory-based recurrent neural network. Experimental results demonstrate that the proposed models achieve more accurate estimation of the ReTM using five objective metrics compared to the covariance-based method. We also show the effectiveness of the proposed frameworks for speech enhancement, achieving performance on par with the baseline method.
Read the full abstract

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

Luna-TTS Family Technical Report

Feng Yin, Shuai Shi, Junjie Zheng et al. (+19 more)

Modern text-to-speech (TTS) is dominated by autoregressive (AR) codec language models, whose left-to-right decoding brings latency that grows with utterance length, error accumulation along the committed prefix, and an artificial generation order imposed on the Residual Vector Quantization (RVQ) token grid. We propose Luna-TTS Family, diffusion-language-model-based TTS systems pretrained on 1 million hours of speech across Chinese, English, Japanese, and Korean. The family is built by progressive adaptation of a pretrained AR text LLM, from causal to bidirectional and finally to block-causal attention, and comprises two variants sharing a single tokenizer, data pipeline, and 0.6B backbone lineage. Luna-TTS is fully non-autoregressive: it generates the entire RVQ token grid in a fixed number of parallel refinement steps, with zero-shot voice cloning and speech editing arising natively as infilling. Luna-TTS Realtime, derived by continual training, is autoregressive over blocks of 32 codec frames (1.28s) while denoising each block in parallel; it supports KV-cached blockwise generation and incremental audio delivery, achieving an end-to-end RTF of 0.0240 and 41.6 ms local first-block latency under the warmed serving protocol. An annealed fine-tuning stage adds explicit control over emotion and non-verbal vocalizations (NVVs), and a reinforcement-learning stage applies GRPO with policy ratios computed over the realized denoising trajectory. On Seed-TTS-Eval, Luna-TTS achieves the best results on all four metrics among compared open-source and commercial systems (0.73 CER / 79.7 SIM on test-zh, 1.49 WER / 76.8 SIM on test-en); on the harder in-the-wild CV3-Eval, it posts the lowest Mandarin and English error rates in our comparison. Against leading commercial systems, it achieves the best results on most objective, model-based, and human-rated metrics for NVV and emotion control.
Read the full abstract

All authors Feng Yin, Shuai Shi, Junjie Zheng, Kechenying Zhou, Yiqiu Wang, Chenyang He, Qiuhua Jiang, Mengxiao Bi, Yanmin Qian, Mingxin Chen, Xun Gong, Tianteng Gu, Bing Han, Peng Jiang, Chenda Li, Haiyang Sun, Han Wang, Wei Wang, Yi Wang, Leying Zhang, Wangyou Zhang, Chushu Zhou

Audio and Speech Processing

Robust Multi-Tier Infant-Centered Audio Understanding with Whisper via Structured Speaker Conditioning

Xulin Fan, Jialu Li, Mohammad Nur Hossain Khan et al. (+4 more)

Recent advances in model design and self-supervised audio representations have improved speech and audio understanding, yet infant-centered naturalistic recordings remain challenging due to limited labeled data, low signal-to-noise ratio, and cross-family domain shifts. We present a family-conditioned, multi-tier audio tagger that combines a LoRA-finetuned Whisper encoder with a lightweight, target-speaker-aware Transformer for long-context inference and framewise prediction across tiers. To improve temporal coherence, we incorporate a simple sequence-level smoothing loss, and to enhance robustness across households, we introduce a factorized speaker-token design with a shared tier token and a learned family-specific offset, reducing family bias and promoting generalizable representations. Together, these choices enable efficient and effective infant-centered audio tagging of daylong audio recordings in home environments.
Read the full abstract

All authors Xulin Fan, Jialu Li, Mohammad Nur Hossain Khan, Kexin Hu, Bashima Islam, Mark Hasegawa-Johnson, Nancy L. McElwain

Tuesday 11 August

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

TTS-Q: Perception-Guided Post-Training Quantization of Text-to-Speech Diffusion Transformers

Xiaoqin Feng, Yuqi Li, Weilun Feng et al. (+9 more)

Zero-shot text-to-speech (TTS) can clone a voice from seconds of audio. Yet the quality comes from billion-parameter diffusion transformers whose size and multi-step sampling put them out of reach of the on-device, assistive, and real-time applications that need them most. Post-training quantization (PTQ) shrinks such models without retraining, but it has been built for language and vision, not speech. At 4 bits, a strong quantizer leaves intelligibility close to full precision but consistently erodes speaker timbre, so error rates stay low while the voice drifts from the target speaker. We ask why quantization hurts speech, where the damage lives, and whether it can be repaired without retraining. To find out, we run the first 4-bit weight-and-activation (W4A4) PTQ study of a TTS diffusion transformer at 1B and 3.5B. The timbre cost comes from both the weight and the activation quantization, but only the activation part has the timestep structure a precision schedule can exploit. We therefore introduce TTS-Q, which selects hard-to-quantize items from task-matched calibration candidates and applies a budget-balanced mixed-precision schedule that assigns higher precision to late-step feed-forward activations. TTS-Q recovers timbre best at both scales, at no cost to intelligibility. Separately, the deployed calibrated-W4A4 reference reduces peak memory on GPU, providing a practical reference for resource-constrained TTS deployment.
Read the full abstract

All authors Xiaoqin Feng, Yuqi Li, Weilun Feng, Yi-Cheng Lin, Kai Li, Xianglong Wang, Kuo Yang, Chuanguang Yang, Yingli Tian, Wei Li, Zhongwei Cheng, Lin Chen

Audio and Speech Processing

MAJEPPA: Morphing and Assessing in a Unified Piano Performance Space

Jinwen Zhou, Huan Zhang, Weixi Zhai et al. (+3 more)

We present MAJEPPA, a self-supervised framework to learn piano performance representations that span the full skill spectrum, from beginner practice sessions to virtuoso concert recordings. We curate the MAJEPPA dataset, comprising ~4,000 annotated recordings across six expertise levels and six recording contexts. We adapt a single pre-trained MIDI autoregressive model with a joint objective: next-token prediction learns score-conditioned performance generation at various skill levels, while InfoNCE and supervised contrastive losses align abstract score and performance representations in a joint embedding space. The proposed model both generates and understands performances in a unified framework. By introducing the EVPMR benchmark, a suite of downstream tasks spanning quality assessment, competition ranking, mistake and technique classification, we evaluate the learnt representations, demonstrating progress towards a real-world model for the piano performance space.
Read the full abstract

All authors Jinwen Zhou, Huan Zhang, Weixi Zhai, Jinhua Liang, Aidan O. T. Hogg, Simon Dixon

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

X2-Turn: Frame-Synchronous Dual-Head Modeling for Joint Streaming ASR and Turn State Prediction

Kaiqi Fu, Rime Wen, Altman Lin et al. (+4 more)

Accurate and responsive turn-taking is essential for spoken dialogue systems, which must distinguish in real time between user interruptions, backchannels that should be ignored, and the completion of an utterance. Prior modular approaches typically optimize turn state prediction at the utterance or fixed-chunk level, creating a mismatch with the continuous turn state estimate, and often depend on an auxiliary ASR model, which limits responsiveness and increases overall system complexity. Therefore, we present X2-Turn, a frame-synchronous turn state prediction method via delayed-stream modeling. Specifically, building on the pretrained Voxtral Realtime model, we introduce a frame-synchronous turn state head that operates in parallel with the ASR head on shared streaming representations, jointly predicting ASR tokens and fine-grained turn states at the frame level. Experiments on bilingual EasyTurn and Full-Duplex-Bench demonstrate that the proposed method achieves an effective trade-off between turn state accuracy and decision latency.
Read the full abstract

All authors Kaiqi Fu, Rime Wen, Altman Lin, Shawn Qin, Roy Gan, Hao Wang, Qian Wang

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

MazzikaAI: A knowledge-based performance-to-prompt compiler for real-time Arabic maqam accompaniment with a streaming text-to-music model

Jiaxin Du, Boulbaba Abdeljaouad, Yong Zhuang et al. (+1 more)

Arabic maqam music microtonal, modal, and built on ornamented call and response is among the traditions most underserved by generative music models, whose training frameworks remain predominantly Western and equaltempered. Real time accompaniment sharpens this gap: an AI partner must listen, adapt dynamically, and respect idiomatic microtonal structures. Streaming text to music models provide strong generative capabilities but lack precise control interfaces. We present MazzikaAI, a knowledge based system that uses natural language as the actuator of a realtime control loop. By compiling live MIDI, gesture, and inferred harmony into continuously updated text prompts, MazzikaAI steers an unmodified streaming generator, Google Lyria RealTime, without requiring model finetuning. The system embeds expert knowledge of six core maqamat, characteristic ornaments, and ensemble dynamics, maintaining realtime responsiveness with subsecond keytoaudibleupdate latency. Empirical evaluations demonstrate that dynamic prompt compilation reliably grounds generation in microtonal scales, significantly increasing offgrid quartertone content over baseline generation. Beyond its core implementation, MazzikaAI illustrates how deterministic knowledgebased rules can effectively bridge expert, nonWestern musical traditions and unfinetuned foundation models. This architecture establishes a scalable paradigm for realtime humanAI cocreation, offering a generalizable blueprint for interactive accompaniment, adaptive music education, and culturally inclusive generative audio across diverse global idioms.
Read the full abstract

All authors Jiaxin Du, Boulbaba Abdeljaouad, Yong Zhuang, Haoyu Li

Monday 10 August

Audio and Speech Processing

A Hybrid Classical-Learning Framework for Adaptive Decision Directed Speech Enhancement

Ali Rajabi and Xiangwei Zhou

Speech enhancement aims to recover clean speech signals from noisy observations while preserving speech quality and intelligibility. Classical methods such as Spectral Subtraction and Decision-Directed (DD) enhancement remain widely used because of their interpretability and low computational complexity, but they may suffer from musical-noise artifacts or excessive attenuation of weak speech components under low signal-to-noise ratio (SNR) conditions. This paper proposes an Adaptive Beta-Constrained Decision-Directed (ABCDD) speech enhancement framework that extends the conventional DD method through a frame-dependent lower gain bound. The introduced beta parameter controls the tradeoff between noise suppression and speech preservation. To automate parameter selection for large and diverse datasets, a lightweight multilayer perceptron (MLP) model is further developed to predict frame-level beta values directly from noisy-speech features. The proposed framework is evaluated using both a representative speech example and large-scale testing on the VoiceBank-DEMAND dataset. In the representative example, ABCDD outperformed conventional Spectral Subtraction and classical DD across multiple objective metrics, including SNR, Log-Spectral Distance (LSD), Root-Mean-Square Error (RMSE), correlation, and Scale-Invariant Signal-to-Distortion Ratio (SI-SDR). On 100 unseen VoiceBank-DEMAND test files, the proposed MLP-beta ABCDD method improved average scale-aligned SNR from 9.41 dB to 13.82 dB, corresponding to an average gain of 4.41 dB. The results indicate that combining interpretable classical enhancement structure with lightweight machine-learning-based parameter adaptation provides an effective and practical direction for robust speech enhancement.
Read the full abstract

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

Beyond the Raw Waveform: Fusing Visual Representations of EDA for Stress Detection

Stefanos Gkikas, Thomas Kassiotis, Yang Guo et al. (+5 more)

Electrodermal activity (EDA) is widely used in automatic stress detection, yet most pipelines treat it only as a raw one-dimensional waveform. This study examines whether complementary visual representations of EDA provide useful information for stress classification and whether their fusion im- proves recognition performance. Six image-based representations are derived from each EDA recording: an unwrapped short-time Fourier transform (STFT) phase spectrogram, an instantaneous-frequency map computed from that phase, a power spectral density (PSD) spectrogram, a continuous wavelet transform scalogram, a recurrence plot, and a rendered waveform trace. The selected representations are stacked as channels of a single multichannel input, together with the raw waveform, and processed by a shared asymmetric-attention architecture. Experiments on a 58-subject stress dataset show that representation fusion improves over the raw waveform. The best configuration, which combines five representations while excluding the unwrapped phase spectrogram, reaches 70.97% test accuracy, compared with 67.36% for the raw waveform. The single PSD spectrogram achieves 69.44%, remaining close to the best-fused configuration at a lower computational cost. The results show that alternative visual forms of the same EDA signal can provide useful inductive biases for stress detection, and that a compact selection of complementary representations can be more effective than the raw waveform alone.
Read the full abstract

All authors Stefanos Gkikas, Thomas Kassiotis, Yang Guo, Guangliang Li, Eric Nichols, Houshyar Asadi, Nikolaos Smyrnis, Giorgos Giannakakis

Audio and Speech Processing

In Defense of Using Worst-case Privacy Disclosure as Privacy Evaluation Metric of Voice Anonymization

Xin Wang and Xiaoxiao Miao

The voice anonymization community mainly uses Equal Error Rate (EER) to evaluate the performance of voice identity protection. While alternative metrics such as privacy-ZEBRA and a rank-based metric have been proposed, their underlying assumptions and differences may not be well known, especially to newcomers. This paper is motivated to fill the gap. Based on the concept of Shannon's perfect secrecy (or privacy), this paper positions itself as a defense of the privacy-ZEBRA framework. While no new metric is proposed, this paper explains how an `ideal' system in terms of EER may fail to gauge the information leakage on individual speakers in the log-likelihood ratio (LLR) space. The paper also shows how the rank-based metric can be cast into a metric that follows the same principle of perfect secrecy and how their best solutions are equivalent. Furthermore, the paper explains how the method of estimating LLRs may affect the evaluation results. These discussions are, to the best of the authors' knowledge, not explored or explained in detail in existing papers. Last but not least, the findings are demonstrated on simulated and VoicePrivacy Challenge data.
Read the full abstract

2608.10318 · 10 Aug · Workshop version accepted by SPSC 2026. Notebook: https://github.com/nii-yamagishilab/paper-archive-spsc2026-privacy-llr. Acknowledgement: we thank the reviewers for the comments; we addressed many of them, and some important comments have to be left to future work in the form of a more comprehensive paper. V2 fixed a typo in Eq.9 \arg\max -> \max

Abstract PDF $PRIV ×5

Audio and Speech Processing

BiTSE: Binaural Target Speaker Extraction in Noisy Multi-Talker Environments for AR Glass Arrays

Selani A. Indrapala and Wageesha N. Manamperi

Isolating a desired speech signal in noisy multi-talker conversational scenarios is a key requirement for augmented reality (AR) wearable microphone array systems. In this work, a binaural target speaker extraction (TSE) framework, termed BiTSE, is proposed. It leverages both spatial and temporal cues, specifically the direction-of-arrival (DoA) of the target speaker and corresponding voice activity information, to guide the extraction process. Built upon a binaural signal denoising architecture, our model integrates three key enhancements: (i) a DoA-aware attention mechanism using cyclic positional embeddings, (ii) a timestamp-based masking strategy that utilizes speaker activity to suppress non-target segments, and (iii) a novel two-stage loss optimization strategy that first trains the model for robust denoising and then fine-tunes it to improve perceptual quality. Evaluations on the SPeech Enhancement for Augmented Reality (SPEAR) challenge dataset demonstrate that the proposed BiTSE consistently improves upon conventional approaches, leading to enhanced signal fidelity and perceptual quality.
Read the full abstract

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

Training Set Synthesis for Bioacoustic Denoising: A Case Study With Mice

Reyhaneh Abbasi, Peter Balazs, Vincent Lostanlen et al. (+4 more)

Bioacoustic recordings are often degraded by ambient noise, which complicates the analysis of weak or noise-overlapped vocalizations. Convolutional neural networks, particularly U-Net architectures, have shown a strong denoising performance in speech and music processing. However, their direct application to bioacoustic signals is limited by the scarcity of clean training data. To address this issue, we propose a training set synthesis approach and develop a supervised denoising model that predicts a complex ratio mask in the time-frequency domain. The model leverages ridges, or frequency contours, that represent the fundamental frequency together with one or more harmonic partial components of vocalizations. These ridges are used both for the synthesis of training sets and to design a loss function that assigns higher weights to the ridge regions (ridge-guided loss function). This weighting step helps the network better preserve vocalization details during denoising. As a case study, we evaluate our approach using ultrasonic vocalizations (USVs) recordings of house mice, which are widely studied in behavioral biology and neuroscience. In actual field recordings, the proposed method enhances fundamental and harmonic partial ridge tracking compared to our previous signal-processing approach. In addition, a classifier trained on denoised data improves USV classification on out-of-sample, noisy recordings from wild and domesticated mice compared to classifiers trained on noisy recordings. Our proposed method also substantially improves the scale-invariant signal-to-distortion ratio on synthetic testing data across a wide range of input signal-to-noise ratios. Although we focus on USVs, the proposed approach should be broadly applicable to other bioacoustic signals with trackable ridges, and thus enables ridgebased training set synthesis and denoising.
Read the full abstract

All authors Reyhaneh Abbasi, Peter Balazs, Vincent Lostanlen, Clara Hollomey, Dustin J. Penn, Sarah M. Zala, Nicki Holighaus

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

Structured Phonological Representations for Audio-Articulatory rtMRI Speech Classification

Abner Hernandez, Tomás Arias Vergara, Daiqi Liu et al. (+2 more)

Real-time MRI makes it possible to observe vocal-tract articulation during speech, but mapping these articulatory patterns to phonetic and phonological categories remains challenging. We investigate whether PhonoQ, an audio-based model trained to recognize structured phonological features, provides useful information for audio–articulatory modeling. Specifically, we extract representations from PhonoQ's Conformer module, whose training is shaped by supervision for manner, place, voicing, and vowel features. Using articulatory contours with synchronized audio-derived features, we compare WavLM-large and HuBERT-large baselines with models that incorporate PhonoQ-derived representations. Across unseen-speech and unseen-subject settings, these features improve macro-F1 for phonological targets including manner, place, voicing, vowel height, and vowel backness, and also improve fine-grained 39-phoneme classification. In a contour-only inference setting, audio-derived teacher supervision yields modest but consistent gains over contour-only training, indicating that phonological information from synchronized audio can be partially transferred to articulatory models. Finally, posterior analyses show interpretable surface-sensitive patterns consistent with flapping-like /t/ realizations, /t/-/r/ retraction or affrication, and nasal place assimilation.
Read the full abstract

All authors Abner Hernandez, Tomás Arias Vergara, Daiqi Liu, Andreas Maier, Paula Andrea Pérez-Toro

Audio and Speech Processing

Dynamic Clustering for Cross-Segment Permutation Alignment in Long Speech Separation

Yuzhu Wang, Archontis Politis, Konstantinos Drossos et al. (+1 more)

Long speech separation typically employs a segment-separation-stitch paradigm where recordings are divided into short segments, processed independently, and stitched together. Its challenge lies in predicting cross-segment permutations. This paper proposes a training-free dynamic clustering approach for cross-segment permutation alignment using speaker embedding reference pools. The method predicts the permutation using the cosine similarity between current segment embeddings and the reference pools. The approach updates reference pools by retaining the most representative speaker embeddings based on their overall cosine similarity with existing references. As a plug-and-play post-processing module compatible with existing separation models, the proposed method demonstrates superior performance compared to existing methods on dense and sparse long speech scenarios, particularly in challenging sparse scenarios with extended utterance gaps, and further shows robustness to speaker count estimation errors in unknown speaker count scenarios.
Read the full abstract

All authors Yuzhu Wang, Archontis Politis, Konstantinos Drossos, Tuomas Virtanen

Audio and Speech Processing

Neural Array-Generic Direction-of-Arrival Estimation Exploiting Array Transfer Functions

Mikko Heikkinen, Archontis Politis, Konstantinos Drossos et al. (+1 more)

Direction-of-arrival (DoA) estimation is a key component of multichannel audio processing, yet many deep learning approaches remain tied to the microphone arrays used during training and generalize poorly to unseen devices. This paper proposes an array-generic neural DoA estimation framework using measured or simulated complex directional array transfer functions (ATFs) matched to real-world multi-microphone devices. The method processes multichannel spectrograms and ATF metadata with separate convolutional encoders, fuses the resulting representations through cross-attention, and predicts source directions using a multi-source Cartesian vector output formulation. Experiments on simulated 2D and 3D localization tasks under reverberation and diffuse babble noise show that the proposed approach generalizes to previously unseen arrays, including mobile-phone-like configurations, without major performance degradation, while remaining competitive with conventional and learning-based baselines.
Read the full abstract

All authors Mikko Heikkinen, Archontis Politis, Konstantinos Drossos, Tuomas Virtanen

Audio and Speech Processing

Speaker Role and Language Diarization for Analyzing Multilingual Interviews for Language Proficiency of Older Adults

Anfeng Xu, Tiantian Feng, Kevin Huang et al. (+8 more)

Automatic language proficiency assessment in the context of multilingual interview-based settings remains underexplored. In this work, we develop Whisper-based speaker-role and language diarization systems to automatically extract respondent speech and characterize language usage in multilingual interviews with older adults. We further investigate whether diarization-derived conversational and language-use behaviors can support downstream language proficiency assessment. Results show that language-adapted Whisper models substantially improve language diarization performance for lower-resource and linguistically related Indian languages. Statistical analyses reveal that respondent speech ratio and intended language usage are strong predictors of proficiency ratings. Furthermore, simple diarization-derived behavioral features achieve performance comparable to Whisper-based speech embeddings for proficiency prediction, while combining both yields the best results. Importantly, both the speech and language use statistical analyses and language proficiency prediction performance remain largely preserved when using fully automatic diarization outputs, demonstrating the potential of respondent-centric conversational analysis for scalable language proficiency assessment.
Read the full abstract

All authors Anfeng Xu, Tiantian Feng, Kevin Huang, Pranali Khobragade, Sudarsana Kadiri, Anushikha Dhankhar, Madeleine Snider, Sarah Gao, Miguel Arce Rentería, Jinkook Lee, Shrikanth Narayanan

Audio and Speech Processing

Beyond Piano: Cross-Instrument MIDI Velocity Estimation via Differentiable SoundFont Proxies

Zhanhong He, Hanyu Meng, David Defeng Huang et al. (+1 more)

Many music datasets contain MIDI notes but lack reliable velocities, defaulting to a constant value. This absence is especially problematic outside the piano domain, as velocity is a core component for expressive rendering, music generation, and performance analysis. This paper studies cross-instrument MIDI velocity estimation in this label-scarce setting. Starting from a piano-trained velocity estimator, we recast target-instrument adaptation as predicting renderer-conditioned velocities whose rendering matches the dynamics of the performance audio. This adaptation can be driven by either differentiable synthesizers (Diff-Synth) or our proposed differentiable SoundFont proxies (Diff-SFProxy). We highlight the Diff-SFProxy: it supervises velocity through note-wise, loudness-related acoustic parameters rather than waveform reconstruction, focusing gradients on velocity-dependent behavior. Experiments on piano and guitar show that Diff-SFProxy is effective for cross-instrument MIDI velocity estimation, while waveform-domain Diff-Synth degrades performance.
Read the full abstract

All authors Zhanhong He, Hanyu Meng, David Defeng Huang, Roberto Togneri

Sunday 9 August

Audio and Speech Processing

Beyond Reconstruction: Full-Context Generative DiT for Music Generation

Yunjia Li, Menglin Wu, Junyu Dai et al. (+13 more)

Hybrid music generators combine the long-range planning of an autoregressive language model with the fidelity of a diffusion- or flow-based acoustic renderer. Yet renderers are trained with clean, target-derived codec tokens but deployed with imperfect language-model predictions, creating codecinterface exposure bias. Rather than treating rendering as a simple reconstruction task,we formulate it as full-context generation from an imperfect discrete plan. We introduce FullDiT, a conditional DiT that fuses eight frame-aligned RVQ streams with independently encoded captions and lyrics and uses non-causal self-attention over the complete acoustic latent sequence. During training, Error-Matched Distractor Conditioning (EMDC) matches per-codebook replacement rates to teacher-forced top-1 error rates and samples near-miss tokens from cosine-KNN neighborhoods without changing the acoustic target. At inference, four-way classifier-free guidance (4-CFG) independently scales codec, lyric, and caption guidance increments. Matched ablations show that EMDC improves ViSQOL by 0.77 under synthetic corruption and is clearly preferred in non-tied comparisons with fixed languagemodel tokens. Further ablations show gains from full-song context and renderer-side text conditioning. The complete system outperforms five commercial systems on 15 of 18 automatic metrics and ranks among the top three on the Artificial Analysis Music with Vocals Leaderboard. The demo page is available at https://selinacloudl.github.io/fulldit-demo/.
Read the full abstract

All authors Yunjia Li, Menglin Wu, Junyu Dai, Xinyue Fan, Xiangang Li, Haoxu Wang, Jianwei Yu, Huaicheng Zhang, Han Zhao, Weiqin Li, Yufei Shi, Cheng Wen, Sitong Zhao, Qixi Zheng, Haina Zhu, Wei Li

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

Multilingual Emotion Neurons in Large Audio-Language Models

Xiutian Zhao, Philipp Koehn, Björn Schuller et al. (+1 more)

Emotion is central to human communication, and its expression varies across languages. Large audio-language models (LALMs) achieve strong performance on multilingual speech tasks, yet it remains unclear whether they encode emotion through language-specific correlations or language-agnostic representations. We present the first neuron-level interpretability study of this question. We define Multilingual Emotion Neurons (MLENs) as functional units exhibiting stable emotional selectivity and aligned causal effects across languages, and introduce Consistency-Regularized Fusion (CR-Fusion) to identify them. Across four modern LALMs and 12 typologically diverse languages, emotion-sensitive neurons identified independently per language show minimal overlap, and additional monolingual identification data saturates quickly without isolating more transferable units, motivating identification from pooled cross-lingual evidence. Causal interventions demonstrate that MLENs identified by CR-Fusion provide more precise and transferable affective control than monolingual neuron sets in both zero-shot and low-resource settings. Leave-one-out ablations further reveal asymmetric transfer: individual identification languages, including low-resource ones, contribute non-redundant evidence, while several low-resource languages benefit most from the resulting cross-lingual transfer. Together, our findings provide the first causal, neuron-level account of how LALMs encode emotion across languages, and establish multilingual neuron identification as an effective mechanism for understanding cross-lingual affective behavior.
Read the full abstract

All authors Xiutian Zhao, Philipp Koehn, Björn Schuller, Berrak Sisman

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

Steering dense music retrieval with open-vocabulary concept discovery

Julien Guinot, Alain Riou, Elio Quinton et al. (+1 more)

Controllable music retrieval lets users find music that is, for example, more ambient, less distorted, or without guitar while preserving the other semantic content of an original seed query. Sparse autoencoders (SAEs) are a promising interface for this kind of concept-level control, but a key problem remains: given a free-form text concept, which sparse features should be edited? In shared multimodal embedding spaces, standard attribution methods often select neurons that match the concept's wording but not the audio examples that express it. This leads to weak or unstable edits: relevant features are missed when concepts are distributed across neurons, while others are selected due to text alignment rather than audio-side structure. We address this with a lightweight, training-free method that recovers a sparse set of audio features whose decoded representation reconstructs the target concept while remaining consistent with audio-space geometry. This reframes concept attribution as a sparse inversion problem rather than a text-side neuron-ranking heuristic. The method requires neither paired audio-text supervision nor SAE retraining. We evaluate this approach in steerable music retrieval and show that the recovered supports align more closely with concept-bearing audio examples and achieve a stronger trade-off between edit strength and preservation than alignment baselines, enabling more precise concept amplification and suppression with reduced drift on preservation metrics.
Read the full abstract

All authors Julien Guinot, Alain Riou, Elio Quinton, György Fazekas

Audio and Speech Processing

A Unifying Perspective on Audio Generative Modeling: Latent Representations and Modeling Strategies

Dongchao Yang

Every audio generative system makes two coupled decisions: what representation to generate, and how to model its distribution. This paper organizes audio generative modeling around this coupling. For representation design, we compare discrete, continuous, and hybrid latents through four objectives: representation burden, distortion, empirical modelability, and streaming compatibility. For distribution modeling, rather than treating a latent's difficulty as an intrinsic scalar, we use two diagnostic dimensions: dependency horizon, how far useful context extends, and conditional ambiguity, how much uncertainty remains after conditioning. These dimensions refine the common semantic-versus-acoustic intuition: variables with a long dependency horizon should receive global modeling capacity. Conditionally ambiguous detail may be delegated to a local or iterative generator. Applied to representative systems, this view shows that RVQ's residual order gives ordered capacity but not ordered semantics, that AudioLM's semantic-versus-acoustic cascade is one explicit placement of this boundary rather than a universal template, and that autoregression, iterative refinement, and hybrid designs differ chiefly in how they trade dependency horizon against critical-path generation cost. The distinction between discrete and continuous latents describes the output interface; dependency horizon, conditional ambiguity, and streaming determine how that interface should be modeled. Rather than cataloguing individual systems, we provide an evaluation and design framework for comparing representation-model pairs.
Read the full abstract

Audio and Speech Processing

BAMU: Bitstream-Aware Marginal-Utility Allocation for Frozen Pretrained Neural Speech Codecs

Mingyu Zhao, Zijian Lin, Yutang Feng et al. (+6 more)

Pretrained neural speech codecs typically use a fixed residual vector quantization (RVQ) depth for all frames, ignoring temporal variation in quantization difficulty. We propose BAMU, a bitstream-aware dynamic RVQ allocation framework for frozen pretrained codecs. A lightweight, rate-independent predictor estimates frame- and layer-wise marginal latent-distortion reductions, while a constrained allocator selects prefix-valid depths under an exact serialized-size budget. Experiments on EnCodec and DAC over LibriSpeech, together with VCTK evaluation, show consistent EnCodec gains and DAC improvements mainly at medium and high rates. A 30-listener study confirms a MOS improvement from 3.449 to 3.780 over matched fixed-depth coding.
Read the full abstract

All authors Mingyu Zhao, Zijian Lin, Yutang Feng, Jiatao Chen, Fan Wang, Jiehui Luo, Yuhao Ding, Jinchao Zhang, Zhiyong Wu

Saturday 8 August

Audio and Speech Processing

CtrlSpeech: Coarse-to-Fine Control for Expressive Speech Synthesis

Zhisheng Zheng, Xiaohang Sun, Zhu Liu et al. (+5 more)

Recent Text-To-Speech (TTS) systems have achieved strong naturalness and zero-shot voice cloning performance, but fine-grained control of expressive speech at the word or phoneme level remains challenging. We propose CtrlSpeech, a controllable, expressive TTS framework with coarse-to-fine control. Built on the DiTAR architecture, CtrlSpeech combines global speaker conditioning with phone-aligned pitch, loudness, and duration signals, enabling localized prosodic control while preserving the target speaker's timbre. This design allows users to adjust expressive attributes at a fine temporal granularity, making speech refinement more flexible and controllable. Experimental results show that CtrlSpeech achieves competitive zero-shot TTS performance and improves controllability over expressive attributes, demonstrating its effectiveness for flexible and practical expressive speech synthesis.
Read the full abstract

All authors Zhisheng Zheng, Xiaohang Sun, Zhu Liu, Caren Chen, Rohith Kumar, Manoj Aggarwal, Gerard Medioni, David Harwath

Audio and Speech Processing

ReLMCodec: Designing Predictable Speech Tokens from Pre-Quantization Phoneme Structure

Zixiang Wan, Xusheng Yang, Zheng Wang et al. (+1 more)

Neural speech codecs face a fundamental tension in the language-model era: tokens that support high-fidelity reconstruction are not necessarily easy for autoregressive models to predict. Our controlled analysis of diverse codec and self-supervised speech representations shows that clearer phoneme structure before discrete code assignment is consistently associated with easier autoregressive token prediction. Yet phoneme structure alone is insufficient for high-fidelity reconstruction, which also requires reconstruction-relevant acoustic detail. Guided by this observation, we introduce ReLMCodec, a low-bitrate single-codebook speech codec built upon a preserve–control–refine principle: it preserves the linguistic organization of frozen self-supervised learning (SSL) features at the quantizer input, controls reconstruction-driven drift through Pre-quantization Anchor-Preserving Adaptation (PAPA), and refines the quantized latent space with a training-only WavLM-Large L24 teacher to reduce phoneme-level token fragmentation. Together, these components allow acoustic detail to support waveform reconstruction while keeping the resulting token sequence predictable for autoregressive models. At 650 and 800 bps, ReLMCodec moves the empirical single-stream predictability–reconstruction frontier in our evaluations, with gains that carry over to downstream text-to-speech (TTS) synthesis in both intelligibility and speaker similarity.
Read the full abstract

All authors Zixiang Wan, Xusheng Yang, Zheng Wang, Peiji Yang

Audio and Speech Processing

SraVaani 1.0: Scaling Inclusive Speech Recognition for Indic Languages

Sujith Pulikodan, Agneedh Basu, Pavan Kumar J et al. (+4 more)

India's linguistic landscape spans over 700 languages and thousands of dialects, yet the vast majority of automatic speech recognition (ASR) systems support only a small fraction of this diversity. We present SraVaani-1.0, a multilingual ASR model covering 65 Indian languages and dialects, many of which currently have no publicly available or competing ASR system. SraVaani-1.0 is built on a FastConformer architecture and trained from scratch through a three stage the first stage, we perform self-supervised pretraining on 31,255 hours of unlabelled speech from the VAANI corpus using a contrastive learning objective. In the second stage, we introduce an audio-image representation alignment stage that leverages the paired images and speech available in the VAANI corpus. This multimodal alignment encourages the speech encoder to learn semantically richer representations by exploiting the relationship between visual context and spoken content, thereby improving downstream recognition, particularly for low resource the final stage, the aligned encoder is fine-tuned end-to-end using a Hybrid Token-and-Duration Transducer (TDT)-CTC decoder on 31,263 hours of labelled multilingual Indian speech compiled from 24 public datasets spanning 65 languages and dialects. We evaluate SraVaani-1.0 against three state-of-the-art multilingual ASR systems across eight benchmarks. SraVaani-1.0 achieves the lowest word error rate (WER) on a large number of language-dataset pairs while remaining competitive with the best-performing systems on high resource importantly, it is the only open-source evaluated model that provides transcription capability for multiple low-resource and tribal Indian languages, which are assessed exclusively on the VAANI benchmark.
Read the full abstract

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

Audio and Speech Processing

The Voiceprint Fallacy: Why Voices Are Not Unique Biometric Imprints

Tianle Yang, Cuiling Zhang, Chengzhe Sun et al. (+2 more)

In recent years, the term voiceprint has regained attention, particularly in technological applications and policy-making contexts, often carrying the assumption that a person's voice constitutes a stable and unique biometric trace analogous to a fingerprint. Yet this conception has been repeatedly criticized and rejected by forensic voice experts throughout the decades since its introduction. Although voices undoubtedly contain speaker-related information, this simplified conception obscures the highly dynamic and context-dependent nature of speech. This article revisits the voiceprint fallacy and reconsiders what can count as evidence of speaker identity by reviewing the historical development of voiceprint identification, evidence on human voice variability, developments in forensic voice comparison, research on human and automatic speaker recognition, and the recent challenge posed by deepfake speech to speaker identity. We point out that the voiceprint metaphor and its underlying implications are scientifically misleading because they transform a probabilistic source of speaker information into an imagined stable object of identity. We argue that speaker identity assessment does not require, and current evidence does not support, the existence of a stable and individually unique voiceprint. For speaker recognition and voice biometrics, this distinction motivates interpreting learned speaker representations with respect to the conditions under which they are trained and evaluated, and explicitly assessing their robustness to relevant sources of within-speaker variability, domain mismatch, and synthetic manipulation.
Read the full abstract

All authors Tianle Yang, Cuiling Zhang, Chengzhe Sun, Siwei Lyu, Phil Rose

Audio and Speech Processing

VIOLET: High-Fidelity Violin Synthesis with Techniques and Dynamics

Baotong Tian, Cynthia Lu, Vincent K.M. Cheung et al. (+3 more)

Neural synthesis for musical instruments has the potential to revolutionize current practices that use concatenative synthesis and a sample library. However, most research focused on piano synthesis and expressive performance generation; little work has been done on continuously articulated instruments like the violin, let alone rendering them with playing techniques and dynamics. We present VIOLET, a latent-diffusion framework for controllable violin synthesis, which uses a Diffusion Transformer (DiT) with rectified flow to synthesize high-fidelity audio from MIDI notes, playing techniques, and continuous dynamics. To train VIOLET, in addition to using a few existing datasets, we curate a new dataset named CSV-TD, which contains 39 h of 48 kHz synthetic audio and time-aligned annotations of MIDI notes, note-level techniques, and continuous dynamics curves. Objective and subjective evaluations show that VIOLET synthesizes violin performances with high technique adherence, accurate pitch and timing alignment, and good dynamics control. It outperforms the current state-of-the-art neural violin synthesis system and approaches a top commercial virtual instrument in terms of technique clarity, naturalness, and dynamics following.
Read the full abstract

All authors Baotong Tian, Cynthia Lu, Vincent K.M. Cheung, Ting-Kang Wang, Jonathan Churchill, Zhiyao Duan

Friday 7 August

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

Enhancing Audio Reasoning via Semantic Summary Prediction

Francesco Bonzi, Pooneh Mousavi, Cem Subakan et al. (+1 more)

Large Audio Language Models (LALMs) perform well on complex question answering but often show a reasoning gap, where explicit Chain-of-Thought (CoT) reduces accuracy compared to direct answers. We hypothesize that long reasoning sequences shift attention away from the audio input. To address this, we propose SPARE (Semantic Prediction for Audio REasoning), which introduces a register token aligned with the final conclusion using a cosine similarity loss with a Sentence-BERT embedding. This conditions the model's latent space with the target semantic goal before reasoning begins. Experiments on MMAU and MMAR with SALMONN show improved zero-shot reasoning and stronger early attention to audio without additional inference cost.
Read the full abstract

All authors Francesco Bonzi, Pooneh Mousavi, Cem Subakan, Mirco Ravanelli