arXiv subjects

Tomoki Toda

Publications and source records attributed to Tomoki Toda.

At least 19 recordsLinked to original sources

arXiv2026

Pseudo-label distillation for discriminative anomalous sound detection

Discriminative anomalous sound detection (ASD) methods train a feature extractor through a classification task using machine-information labels. They then detect anomalies in the resulting feature space based on distances to normal samples. The discriminative feature space effectively captures machine characteristics, leading to high ASD performance. However, this approach benefits from detailed labels, which are costly to obtain. An alternative is a self-supervised learning (SSL)-based label-free approach. This approach directly uses SSL features for ASD and has shown competitive performance. However, SSL models are typically large and computationally expensive. To address these problems, we propose a simple pseudo-label distillation framework. The proposed method generates pseudo labels from SSL features and trains a compact discriminative feature extractor using these pseudo labels. To suppress the effect of noise on pseudo-label generation, we also propose lightweight noise-robust feature transformation (NRFT) methods utilizing a small amount of clean machine-sound data or isolated noise data. We conducted comprehensive evaluations and analyses on the DCASE 2020-2025 Task 2 datasets using four SSL models. The results demonstrate that pseudo-label distillation not only transfers the performance of SSL models to a compact model but also further improves performance by leveraging available coarse labels and data augmentation. Also, our NRFT methods provide further gains.

eess.AS↗

arXiv2026

Attacking UTMOS: Probing the Robustness of a Speech Quality Assessment Model

UTMOS has become one of the most commonly used deep neural network-based speech quality assessment (SQA) metrics in speech processing research. In this paper, we attack UTMOS to probe its robustness. Starting from high-quality speech samples, we optimize the input in two directions: a score-preserving attack, which degrades perceived quality while maintaining the predicted score, and a quality-preserving attack, which lowers the predicted score while maintaining perceived quality. We consider three input spaces: raw waveform, mel spectrogram with a HiFi-GAN vocoder, and the latent space of EnCodec, a neural audio codec. Experimental results show that score-preserving attacks are effective against UTMOS. Although perfect quality-preserving attacks are more difficult, optimization in the EnCodec latent space provides the best chance of success. These results reveal failure modes of UTMOS and highlight the importance of robustness analysis for DNN-based SQA metrics.

cs.SD↗

arXiv2026

Exploring Pre-training Benefits on Phoneme Addition through Fine-tuning in Speech Synthesis

Transfer learning is widely used for low-resource text-to-speech. When the target corpus contains phonemes unseen in pre-training, the model must expand its phoneme inventory during fine-tuning; we call the process "phoneme addition." However, it remains unclear whether the pre-trained ability to generate seen phonemes contributes to this process. This study investigates phoneme addition in two settings: (1) a simulation setup using LLM-generated phoneme-controlled corpora that enables investigation without considering confounding factors, and (2) a real-speech cross-lingual transfer setup (English to Japanese) to validate whether the findings hold in practice. Experiments in both settings showed that while fine-tuning achieved higher naturalness than training from scratch, it required as much or more data to achieve comparable PER for new phonemes. These results indicate that pre-training mainly contributes to naturalness improvement, but offers limited benefit for phoneme addition.

cs.SD↗

arXiv2026

Learning Emotion-discriminative Representations for Zero-Shot Cross-lingual Speech Emotion Recognition

Zero-shot cross-lingual speech emotion recognition (SER) remains challenging due to distribution mismatches across languages and the lack of emotion annotations in target language. Under such conditions, models trained solely on source-language data frequently suffer from degraded generalization when evaluated on unseen target languages. To address this limitation, we propose an emotion-discriminative representation learning method that integrates supervised contrastive learning and speaker adversarial learning. The contrastive learning promotes cross-lingual emotion alignment, while speaker adversarial learning suppresses speaker-related cues to encourage speaker-invariant representations. Experimental results under a zero-shot cross-lingual SER setting demonstrate that the proposed method significantly improves SER performance over conventional training strategies.

cs.SD↗

arXiv2026

Advancing Electrolaryngeal Speech Enhancement Through Speech-Text Representation Learning

Objective: laryngectomees depend on an electromechanical device to generate electrolaryngeal (EL) speech. Compared with normal speech, EL speech suffers from severe distortion, limited phonetic variation, unnatural prosody, and temporal shifts, degrading naturalness and intelligibility. Although sequence-to-sequence (seq2seq) voice conversion (VC) based EL-speech-to-normal-speech conversion (EL2SP) is promising, substantial mismatches between EL and normal speech inevitably cause cumulative mapping errors that limit performance. To address this, we describe a novel representation learning framework integrating speech and text representations to improve mapping and reconstruction quality within a seq2seq VC model. Methods: our methodology comprises two main stages: 1) representation integration and learning, and 2) reconstruction training. A network capable of incorporating auxiliary text information is first constructed with pretrained modules to learn speech--text-based integrated representations. Then, an autoencoder-style reconstruction strategy finalizes EL2SP model to inherit these representations without increasing model complexity. We introduce three fusion strategies including middle-, input-, and hybrid-level fusion strategies that progressively enhance learning. Moreover, besides standard seq2seq VC objectives, an additional reconstruction loss on the integrated representation is introduced to refine representation transfer. Results: experiments under different EL2SP datasets consistently demonstrate that our methods, combined with data augmentations, outperform baselines relying solely on speech representations. Furthermore, progressive improvements with system design depth validate the effectiveness of our methods. Significance: the proposed methods provide an extensible and practical methodology for EL speech enhancement and assistive communication technologies.

eess.AS↗

arXiv2026

PathBench: Speech Intelligibility Benchmark for Automatic Pathological Speech Assessment

Automatic speech intelligibility assessment is crucial for monitoring speech disorders and therapy efficacy. However, existing methods are difficult to compare: research is fragmented across private datasets with inconsistent protocols. We introduce PathBench, a unified benchmark for pathological speech assessment using public datasets. We compare reference-free, reference-text, and reference-audio methods across three protocols (Matched Content, Extended, and Full) representing how a linguist (controlled stimuli) versus machine learning specialist (maximum data) would approach the same data. We establish benchmark baselines across six datasets, enabling systematic evaluation of future methodological advances, and introduce Dual-ASR Articulatory Precision (DArtP), achieving the highest average correlation among reference-free methods.

cs.SD↗

arXiv2026

Towards explainable reference-free speech intelligibility evaluation of people with pathological speech

Objective assessment of speech that reflects meaningful changes in communication is crucial for clinical decision making and reproducible research. While existing objective assessments, particularly reference-based approaches, can capture intelligibility changes, they are often hindered by lack of explainability and the need for labor-intensive manual transcriptions. To address these issues, this work proposes the reference-free, explainable ASR Inconsistency Score. We evaluate this method on pathological speech in Dutch, Spanish and English, and compare its performance to a reference-based Word Error Rate (WER) baseline. Our results demonstrate that the ASR Inconsistency Score achieves a high correlation with expert perceptual ratings, with performance closely matching, and in one case exceeding, a standard reference-based Word Error Rate (WER) baseline.

cs.SD↗

arXiv2025

EM2LDL: A Multilingual Speech Corpus for Mixed Emotion Recognition through Label Distribution Learning

This study introduces EM2LDL, a novel multilingual speech corpus designed to advance mixed emotion recognition through label distribution learning. Addressing the limitations of predominantly monolingual and single-label emotion corpora \textcolor{black}{that restrict linguistic diversity, are unable to model mixed emotions, and lack ecological validity}, EM2LDL comprises expressive utterances in English, Mandarin, and Cantonese, capturing the intra-utterance code-switching prevalent in multilingual regions like Hong Kong and Macao. The corpus integrates spontaneous emotional expressions from online platforms, annotated with fine-grained emotion distributions across 32 categories. Experimental baselines using self-supervised learning models demonstrate robust performance in speaker-independent gender-, age-, and personality-based evaluations, with HuBERT-large-EN achieving optimal results. By incorporating linguistic diversity and ecological validity, EM2LDL enables the exploration of complex emotional dynamics in multilingual settings. This work provides a versatile testbed for developing adaptive, empathetic systems for applications in affective computing, including mental health monitoring and cross-cultural communication. The dataset, annotations, and baseline codes are publicly available at https://github.com/xingfengli/EM2LDL.

cs.CL↗

arXiv2025

XPPG-PCA: Reference-free automatic speech severity evaluation with principal components

Reliably evaluating the severity of a speech pathology is crucial in healthcare. However, the current reliance on expert evaluations by speech-language pathologists presents several challenges: while their assessments are highly skilled, they are also subjective, time-consuming, and costly, which can limit the reproducibility of clinical studies and place a strain on healthcare resources. While automated methods exist, they have significant drawbacks. Reference-based approaches require transcriptions or healthy speech samples, restricting them to read speech and limiting their applicability. Existing reference-free methods are also flawed; supervised models often learn spurious shortcuts from data, while handcrafted features are often unreliable and restricted to specific speech tasks. This paper introduces XPPG-PCA (x-vector phonetic posteriorgram principal component analysis), a novel, unsupervised, reference-free method for speech severity evaluation. Using three Dutch oral cancer datasets, we demonstrate that XPPG-PCA performs comparably to, or exceeds established reference-based methods. Our experiments confirm its robustness against data shortcuts and noise, showing its potential for real-world clinical use. Taken together, our results show that XPPG-PCA provides a robust, generalizable solution for the objective assessment of speech pathology, with the potential to significantly improve the efficiency and reliability of clinical evaluations across a range of disorders. An open-source implementation is available.

cs.SD↗

arXiv2025

Reference-free automatic speech severity evaluation using acoustic unit language modelling

Speech severity evaluation is becoming increasingly important as the economic burden of speech disorders grows. Current speech severity models often struggle with generalization, learning dataset-specific acoustic cues rather than meaningful correlates of speech severity. Furthermore, many models require reference speech or a transcript, limiting their applicability in ecologically valid scenarios, such as spontaneous speech evaluation. Previous research indicated that automatic speech naturalness evaluation scores correlate strongly with severity evaluation scores, leading us to explore a reference-free method, SpeechLMScore, which does not rely on pathological speech data. Additionally, we present the NKI-SpeechRT dataset, based on the NKI-CCRT dataset, to provide a more comprehensive foundation for speech severity evaluation. This study evaluates whether SpeechLMScore outperforms traditional acoustic feature-based approaches and assesses the performance gap between reference-free and reference-based models. Moreover, we examine the impact of noise on these models by utilizing subjective noise ratings in the NKI-SpeechRT dataset. The results demonstrate that SpeechLMScore is robust to noise and offers superior performance compared to traditional approaches.

cs.SD↗

arXiv2025

M4SER: Multimodal, Multirepresentation, Multitask, and Multistrategy Learning for Speech Emotion Recognition

Multimodal speech emotion recognition (SER) has emerged as pivotal for improving human-machine interaction. Researchers are increasingly leveraging both speech and textual information obtained through automatic speech recognition (ASR) to comprehensively recognize emotional states from speakers. Although this approach reduces reliance on human-annotated text data, ASR errors possibly degrade emotion recognition performance. To address this challenge, in our previous work, we introduced two auxiliary tasks, namely, ASR error detection and ASR error correction, and we proposed a novel multimodal fusion (MF) method for learning modality-specific and modality-invariant representations across different modalities. Building on this foundation, in this paper, we introduce two additional training strategies. First, we propose an adversarial network to enhance the diversity of modality-specific representations. Second, we introduce a label-based contrastive learning strategy to better capture emotional features. We refer to our proposed method as M4SER and validate its superiority over state-of-the-art methods through extensive experiments using IEMOCAP and MELD datasets.

cs.HC↗

arXiv2025

An Extensive Analysis of the Singing Voice Conversion Challenge 2025 Evaluation Results

We present a thorough analysis of the findings of the latest iteration of the Singing Voice Conversion Challenge, a scientific event aiming to compare and understand different voice conversion systems in a controlled environment. Compared to previous iterations which solely focused on converting the singer identity, this year we also focused on converting the singing style of the singer. To create a controlled environment and thorough evaluations, we developed a new challenge database, introduced two tasks, open-sourced baselines, and conducted large-scale crowd-sourced listening tests and objective evaluations. The challenge was run for two months and in total we evaluated 33 different systems. The results of the large-scale crowd-sourced listening test showed that top systems had comparable singer identity scores to ground truth samples. However, modeling the singing style and consequently achieving high naturalness still remains a challenge in this task, primarily due to the difficulty in modeling dynamic information in breathy, glissando, and vibrato singing styles. Further analyses of the challenge also discuss the limitations of both the traditional similarity test and the dynamic preference test in evaluating singing style similarity. Moreover, calculating Spearman's rank correlation coefficient shows that dependent objective metrics such as chroma-alignment and non-match metrics such as speaker embeddings are the most correlated to subjective scores, but are still not at a level where it could be considered as a true replacement for subjective scores.

cs.SD↗

arXiv2025

Speaker Privacy and Security in the Big Data Era: Protection and Defense against Deepfake

In the era of big data, remarkable advancements have been achieved in personalized speech generation techniques that utilize speaker attributes, including voice and speaking style, to generate deepfake speech. This has also amplified global security risks from deepfake speech misuse, resulting in considerable societal costs worldwide. To address the security threats posed by deepfake speech, techniques have been developed focusing on both the protection of voice attributes and the defense against deepfake speech. Among them, the voice anonymization technique has been developed to protect voice attributes from extraction for deepfake generation, while deepfake detection and watermarking have been utilized to defend against the misuse of deepfake speech. This paper provides a short and concise overview of the three techniques, describing the methodologies, advancements, and challenges. A comprehensive version, offering additional discussions, will be published in the near future.

eess.AS↗

arXiv2025

Layer-wise Analysis for Quality of Multilingual Synthesized Speech

While supervised quality predictors for synthesized speech have demonstrated strong correlations with human ratings, their requirement for in-domain labeled training data hinders their generalization ability to new domains. Unsupervised approaches based on pretrained self-supervised learning (SSL) based models and automatic speech recognition (ASR) models are a promising alternative; however, little is known about how these models encode information about speech quality. Towards the goal of better understanding how different aspects of speech quality are encoded in a multilingual setting, we present a layer-wise analysis of multilingual pretrained speech models based on reference modeling. We find that features extracted from early SSL layers show correlations with human ratings of synthesized speech, and later layers of ASR models can predict quality of non-neural systems as well as intelligibility. We also demonstrate the importance of using well-matched reference data.

eess.AS↗

arXiv2025

PARCO: Phoneme-Augmented Robust Contextual ASR via Contrastive Entity Disambiguation

Automatic speech recognition (ASR) systems struggle with domain-specific named entities, especially homophones. Contextual ASR improves recognition but often fails to capture fine-grained phoneme variations due to limited entity diversity. Moreover, prior methods treat entities as independent tokens, leading to incomplete multi-token biasing. To address these issues, we propose Phoneme-Augmented Robust Contextual ASR via COntrastive entity disambiguation (PARCO), which integrates phoneme-aware encoding, contrastive entity disambiguation, entity-level supervision, and hierarchical entity filtering. These components enhance phonetic discrimination, ensure complete entity retrieval, and reduce false positives under uncertainty. Experiments show that PARCO achieves CER of 4.22% on Chinese AISHELL-1 and WER of 11.14% on English DATA2 under 1,000 distractors, significantly outperforming baselines. PARCO also demonstrates robust gains on out-of-domain datasets like THCHS-30 and LibriSpeech.

cs.CL↗

arXiv2025

The AudioMOS Challenge 2025

This is the summary paper for the AudioMOS Challenge 2025, the very first challenge for automatic subjective quality prediction for synthetic audio. The challenge consists of three tracks. The first track aims to assess text-to-music samples in terms of overall quality and textual alignment. The second track is based on the four evaluation dimensions of Meta Audiobox Aesthetics, and the test set consists of text-to-speech, text-to-audio, and text-to-music samples. The third track focuses on synthetic speech quality assessment in different sampling rates. The challenge attracted 24 unique teams from both academia and industry, and improvements over the baselines were confirmed. The outcome of this challenge is expected to facilitate development and progress in the field of automatic evaluation for audio generation systems.

cs.SD↗

arXiv2025

Relationship between objective and subjective perceptual measures of speech in individuals with head and neck cancer

Meaningful speech assessment is vital in clinical phonetics and therapy monitoring. This study examined the link between perceptual speech assessments and objective acoustic measures in a large head and neck cancer (HNC) dataset. Trained listeners provided ratings of intelligibility, articulation, voice quality, phonation, speech rate, nasality, and background noise on speech. Strong correlations were found between subjective intelligibility, articulation, and voice quality, likely due to a shared underlying cause of speech symptoms in our speaker population. Objective measures of intelligibility and speech rate aligned with their subjective counterpart. Our results suggest that a single intelligibility measure may be sufficient for the clinical monitoring of speakers treated for HNC using concomitant chemoradiation.

cs.SD↗

arXiv2025

Unifying Listener Scoring Scales: Comparison Learning Framework for Speech Quality Assessment and Continuous Speech Emotion Recognition

Speech Quality Assessment (SQA) and Continuous Speech Emotion Recognition (CSER) are two key tasks in speech technology, both relying on listener ratings. However, these ratings are inherently biased due to individual listener factors. Previous approaches have introduced a mean listener scoring scale and modeled all listener scoring scales in the training set. However, the mean listener approach is prone to distortion from averaging ordinal data, leading to potential biases. Moreover, learning multiple listener scoring scales while inferring based only on the mean listener scale limits effectiveness. In contrast, our method focuses on modeling a unified listener scoring scale, using comparison scores to correctly capture the scoring relationships between utterances. Experimental results show that our method effectively improves prediction performance in both SQA and CSER tasks, proving its effectiveness and robustness.

eess.AS↗