arXiv subjects

Hoirin Kim

Publications and source records attributed to Hoirin Kim.

At least 19 recordsLinked to original sources

arXiv2026

Language Orthogonalization for Zero-Shot Cross-Lingual Audio Deepfake Detection

Audio deepfake detectors need to transfer to languages absent from training, as multilingual speech synthesis outpaces labeled anti-spoofing resources. While detectors increasingly rely on self-supervised speech models (S3Ms), these backbones encode language-dependent structure that confounds spoof cues. We address this confound through language orthogonalization, a target-free ridge map that removes S3M variation projected onto continuous language-identification (LID) embeddings. Across six languages, six S3M backbones, and all Leave-N-Out settings, it consistently reduces EER across unseen languages. Cross-lingual EER correlates with LID-space distance, where orthogonalization yields larger gains for more distant transfers.

eess.AS↗

arXiv2026

VeRe-Flow: Guiding Flow Matching toward Clean Speech via Velocity Contrastive Regularization and Representation Alignment for Noise-Robust Bandwidth Expansion

Noise-robust bandwidth expansion aims to reconstruct high-fidelity wideband speech from noisy low-resolution inputs. While flow matching has shown strong performance in speech generation, accurately recovering clean speech from noisy inputs remains challenging due to the ambiguity of velocity estimation under noise. In this work, we propose VeRe-Flow, a clean-guided flow matching framework that introduces multi-level clean supervision to guide the generative process toward clean speech. At the velocity level, we introduce velocity contrastive regularization, which attracts the predicted velocity toward the clean trajectory while repelling it from noisy trajectories. At the representation level, we incorporate representation alignment that aligns intermediate features with clean self-supervised learning representations. The results demonstrate that the proposed method achieves the lowest LSD and highest DNSMOS OVRL among all baselines, and the highest MOS among generative baselines.

eess.AS↗

arXiv2026

Scaling Self-Supervised Speech Models Uncovers Deep Linguistic Relationships: Evidence from the Pacific Cluster

Similarities between language representations derived from Self-Supervised Speech Models (S3Ms) have been observed to primarily reflect geographic proximity or surface typological similarities driven by recent expansion or contact, potentially missing deeper genealogical signals. We investigate how scaling an S3M-based language identification system from 126 to 4,017 languages reshapes this topology, and find a non-linear effect: phylogenetic recovery stays flat up to the 1K scale, but the 4K model undergoes a qualitative shift, resolving both clear lineages and long-term linguistic contact. Most strikingly, a robust Pacific macro-cluster emerges, grouping genealogically unrelated Papuan, Oceanic, and Australian languages, and we trace its driver to a concentrated encoding that captures shared acoustic signatures such as global energy dynamics. These results suggest that massive S3Ms internalize multiple layers of language history, offering a promising perspective for computational phylogenetics and the study of language contact.

cs.CL↗

arXiv2026

How Far Do SSL Speech Models Listen for Tone? Temporal Focus of Tone Representation under Low-resource Transfer

Lexical tone is central to many languages but remains underexplored in self-supervised learning (SSL) speech models, especially beyond Mandarin. We study four languages with complex and diverse tone systems (Burmese, Thai, Lao, and Vietnamese) to ask how far such models "listen" for tone and how transfer operates in low-resource conditions. As a baseline reference, we estimate the temporal span of tone cues: approximately 100ms (Burmese/Thai) and 180ms (Lao/Vietnamese). Probes and gradient analysis on fine-tuned SSL models reveal that tone transfer varies by downstream task: automatic speech recognition fine-tuning aligns spans with language-specific tone cues, while prosody- and voice-related tasks bias toward overly long spans. These findings indicate that tone transfer is shaped by downstream task, highlighting task effects on temporal focus in tone modeling.

eess.AS↗

arXiv2025

Two Heads Are Better Than One: Audio-Visual Speech Error Correction with Dual Hypotheses

This paper introduces a new paradigm for generative error correction (GER) framework in audio-visual speech recognition (AVSR) that reasons over modality-specific evidences directly in the language space. Our framework, DualHyp, empowers a large language model (LLM) to compose independent N-best hypotheses from separate automatic speech recognition (ASR) and visual speech recognition (VSR) models. To maximize the effectiveness of DualHyp, we further introduce RelPrompt, a noise-aware guidance mechanism that provides modality-grounded prompts to the LLM. RelPrompt offers the temporal reliability of each modality stream, guiding the model to dynamically switch its focus between ASR and VSR hypotheses for an accurate correction. Under various corruption scenarios, our framework attains up to 57.7% error rate gain on the LRS2 benchmark over standard ASR baseline, contrary to single-stream GER approaches that achieve only 10% gain. To facilitate research within our DualHyp framework, we release the code and the dataset comprising ASR and VSR hypotheses at https://github.com/sungnyun/dualhyp.

eess.AS↗

arXiv2025

HuBERT-VIC: Improving Noise-Robust Automatic Speech Recognition of Speech Foundation Model via Variance-Invariance-Covariance Regularization

Noise robustness in speech foundation models (SFMs) has been a critical challenge, as most models are primarily trained on clean data and experience performance degradation when the models are exposed to noisy speech. To address this issue, we propose HuBERT-VIC, a noise-robust SFM with variance, in-variance, and covariance regularization (VICReg) objectives. These objectives adjust the statistics of noisy speech representations, enabling the model to capture diverse acoustic characteristics and improving the generalization ability across different types of noise. When applied to HuBERT, our model shows relative performance improvements of 23.3% on LibriSpeech test-clean and 13.2% on test-other, compared to the baseline model pre-trained on noisy speech.

cs.SD↗

arXiv2025

ParaNoise-SV: Integrated Approach for Noise-Robust Speaker Verification with Parallel Joint Learning of Speech Enhancement and Noise Extraction

Noise-robust speaker verification leverages joint learning of speech enhancement (SE) and speaker verification (SV) to improve robustness. However, prevailing approaches rely on implicit noise suppression, which struggles to separate noise from speaker characteristics as they do not explicitly distinguish noise from speech during training. Although integrating SE and SV helps, it remains limited in handling noise effectively. Meanwhile, recent SE studies suggest that explicitly modeling noise, rather than merely suppressing it, enhances noise resilience. Reflecting this, we propose ParaNoise-SV, with dual U-Nets combining a noise extraction (NE) network and a speech enhancement (SE) network. The NE U-Net explicitly models noise, while the SE U-Net refines speech with guidance from NE through parallel connections, preserving speaker-relevant features. Experimental results show that ParaNoise-SV achieves a relatively 8.4% lower equal error rate (EER) than previous joint SE-SV models.

eess.AS↗

arXiv2025

Improving Cross-Lingual Phonetic Representation of Low-Resource Languages Through Language Similarity Analysis

This paper examines how linguistic similarity affects cross-lingual phonetic representation in speech processing for low-resource languages, emphasizing effective source language selection. Previous cross-lingual research has used various source languages to enhance performance for the target low-resource language without thorough consideration of selection. Our study stands out by providing an in-depth analysis of language selection, supported by a practical approach to assess phonetic proximity among multiple language families. We investigate how within-family similarity impacts performance in multilingual training, which aids in understanding language dynamics. We also evaluate the effect of using phonologically similar languages, regardless of family. For the phoneme recognition task, utilizing phonologically similar languages consistently achieves a relative improvement of 55.6% over monolingual training, even surpassing the performance of a large-scale self-supervised learning model. Multilingual training within the same language family demonstrates that higher phonological similarity enhances performance, while lower similarity results in degraded performance compared to monolingual training.

eess.AS↗

arXiv2024

Learning Video Temporal Dynamics with Cross-Modal Attention for Robust Audio-Visual Speech Recognition

Audio-visual speech recognition (AVSR) aims to transcribe human speech using both audio and video modalities. In practical environments with noise-corrupted audio, the role of video information becomes crucial. However, prior works have primarily focused on enhancing audio features in AVSR, overlooking the importance of video features. In this study, we strengthen the video features by learning three temporal dynamics in video data: context order, playback direction, and the speed of video frames. Cross-modal attention modules are introduced to enrich video features with audio information so that speech variability can be taken into account when training on the video temporal dynamics. Based on our approach, we achieve the state-of-the-art performance on the LRS2 and LRS3 AVSR benchmarks for the noise-dominant settings. Our approach excels in scenarios especially for babble and speech noise, indicating the ability to distinguish the speech signal that should be recognized from lip movements in the video modality. We support the validity of our methodology by offering the ablation experiments for the temporal dynamics losses and the cross-modal attention architecture design.

eess.AS↗

arXiv2024

One-Class Learning with Adaptive Centroid Shift for Audio Deepfake Detection

As speech synthesis systems continue to make remarkable advances in recent years, the importance of robust deepfake detection systems that perform well in unseen systems has grown. In this paper, we propose a novel adaptive centroid shift (ACS) method that updates the centroid representation by continually shifting as the weighted average of bonafide representations. Our approach uses only bonafide samples to define their centroid, which can yield a specialized centroid for one-class learning. Integrating our ACS with one-class learning gathers bonafide representations into a single cluster, forming well-separated embeddings robust to unseen spoofing attacks. Our proposed method achieves an equal error rate (EER) of 2.19% on the ASVspoof 2021 deepfake dataset, outperforming all existing systems. Furthermore, the t-SNE visualization illustrates that our method effectively maps the bonafide embeddings into a single cluster and successfully disentangles the bonafide and spoof classes.

eess.AS↗

arXiv2024

STaR: Distilling Speech Temporal Relation for Lightweight Speech Self-Supervised Learning Models

Albeit great performance of Transformer-based speech selfsupervised learning (SSL) models, their large parameter size and computational cost make them unfavorable to utilize. In this study, we propose to compress the speech SSL models by distilling speech temporal relation (STaR). Unlike previous works that directly match the representation for each speech frame, STaR distillation transfers temporal relation between speech frames, which is more suitable for lightweight student with limited capacity. We explore three STaR distillation objectives and select the best combination as the final STaR loss. Our model distilled from HuBERT BASE achieves an overall score of 79.8 on SUPERB benchmark, the best performance among models with up to 27 million parameters. We show that our method is applicable across different speech SSL models and maintains robust performance with further reduced parameters.

cs.SD↗

arXiv2023

Recycle-and-Distill: Universal Compression Strategy for Transformer-based Speech SSL Models with Attention Map Reusing and Masking Distillation

Transformer-based speech self-supervised learning (SSL) models, such as HuBERT, show surprising performance in various speech processing tasks. However, huge number of parameters in speech SSL models necessitate the compression to a more compact model for wider usage in academia or small companies. In this study, we suggest to reuse attention maps across the Transformer layers, so as to remove key and query parameters while retaining the number of layers. Furthermore, we propose a novel masking distillation strategy to improve the student model's speech representation quality. We extend the distillation loss to utilize both masked and unmasked speech frames to fully leverage the teacher model's high-quality representation. Our universal compression strategy yields the student model that achieves phoneme error rate (PER) of 7.72% and word error rate (WER) of 9.96% on the SUPERB benchmark.

eess.AS↗

arXiv2023

AdaMS: Deep Metric Learning with Adaptive Margin and Adaptive Scale for Acoustic Word Discrimination

Many recent loss functions in deep metric learning are expressed with logarithmic and exponential forms, and they involve margin and scale as essential hyper-parameters. Since each data class has an intrinsic characteristic, several previous works have tried to learn embedding space close to the real distribution by introducing adaptive margins. However, there was no work on adaptive scales at all. We argue that both margin and scale should be adaptively adjustable during the training. In this paper, we propose a method called Adaptive Margin and Scale (AdaMS), where hyper-parameters of margin and scale are replaced with learnable parameters of adaptive margins and adaptive scales for each class. Our method is evaluated on Wall Street Journal dataset, and we achieve outperforming results for word discrimination tasks.

eess.AS↗

arXiv2022

Anti-Spoofing Using Transfer Learning with Variational Information Bottleneck

Recent advances in sophisticated synthetic speech generated from text-to-speech (TTS) or voice conversion (VC) systems cause threats to the existing automatic speaker verification (ASV) systems. Since such synthetic speech is generated from diverse algorithms, generalization ability with using limited training data is indispensable for a robust anti-spoofing system. In this work, we propose a transfer learning scheme based on the wav2vec 2.0 pretrained model with variational information bottleneck (VIB) for speech anti-spoofing task. Evaluation on the ASVspoof 2019 logical access (LA) database shows that our method improves the performance of distinguishing unseen spoofed and genuine speech, outperforming current state-of-the-art anti-spoofing systems. Furthermore, we show that the proposed system improves performance in low-resource and cross-dataset settings of anti-spoofing task significantly, demonstrating that our system is also robust in terms of data size and data distribution.

eess.AS↗

arXiv2022

FitHuBERT: Going Thinner and Deeper for Knowledge Distillation of Speech Self-Supervised Learning

Large-scale speech self-supervised learning (SSL) has emerged to the main field of speech processing, however, the problem of computational cost arising from its vast size makes a high entry barrier to academia. In addition, existing distillation techniques of speech SSL models compress the model by reducing layers, which induces performance degradation in linguistic pattern recognition tasks such as phoneme recognition (PR). In this paper, we propose FitHuBERT, which makes thinner in dimension throughout almost all model components and deeper in layer compared to prior speech SSL distillation works. Moreover, we employ a time-reduction layer to speed up inference time and propose a method of hint-based distillation for less performance degradation. Our method reduces the model to 23.8% in size and 35.9% in inference time compared to HuBERT. Also, we achieve 12.1% word error rate and 13.3% phoneme error rate on the SUPERB benchmark which is superior than prior work.

eess.AS↗

arXiv2022

Asymmetric Proxy Loss for Multi-View Acoustic Word Embeddings

Acoustic word embeddings (AWEs) are discriminative representations of speech segments, and learned embedding space reflects the phonetic similarity between words. With multi-view learning, where text labels are considered as supplementary input, AWEs are jointly trained with acoustically grounded word embeddings (AGWEs). In this paper, we expand the multi-view approach into a proxy-based framework for deep metric learning by equating AGWEs with proxies. A simple modification in computing the similarity matrix allows the general pair weighting to formulate the data-to-proxy relationship. Under the systematized framework, we propose an asymmetric-proxy loss that combines different parts of loss functions asymmetrically while keeping their merits. It follows the assumptions that the optimal function for anchor-positive pairs may differ from one for anchor-negative pairs, and a proxy may have a different impact when it substitutes for different positions in the triplet. We present comparative experiments with various proxy-based losses including our asymmetric-proxy loss, and evaluate AWEs and AGWEs for word discrimination tasks on WSJ corpus. The results demonstrate the effectiveness of the proposed method.

eess.AS↗

arXiv2022

Learning to Maximize Speech Quality Directly Using MOS Prediction for Neural Text-to-Speech

Although recent neural text-to-speech (TTS) systems have achieved high-quality speech synthesis, there are cases where a TTS system generates low-quality speech, mainly caused by limited training data or information loss during knowledge distillation. Therefore, we propose a novel method to improve speech quality by training a TTS model under the supervision of perceptual loss, which measures the distance between the maximum possible speech quality score and the predicted one. We first pre-train a mean opinion score (MOS) prediction model and then train a TTS model to maximize the MOS of synthesized speech using the pre-trained MOS prediction model. The proposed method can be applied independently regardless of the TTS model architecture or the cause of speech quality degradation and efficiently without increasing the inference time or model complexity. The evaluation results for the MOS and phone error rate demonstrate that our proposed approach improves previous models in terms of both naturalness and intelligibility.

eess.AS↗

arXiv2020

Dynamic Noise Embedding: Noise Aware Training and Adaptation for Speech Enhancement

Estimating noise information exactly is crucial for noise aware training in speech applications including speech enhancement (SE) which is our focus in this paper. To estimate noise-only frames, we employ voice activity detection (VAD) to detect non-speech frames by applying optimal threshold on speech posterior. Here, the non-speech frames can be regarded as noise-only frames in noisy signal. These estimated frames are used to extract noise embedding, named dynamic noise embedding (DNE), which is useful for an SE module to capture the characteristic of background noise. The DNE is extracted by a simple neural network, and the SE module with the DNE can be jointly trained to be adaptive to the environment. Experiments are conducted on TIMIT dataset for single-channel denoising task and U-Net is used as a backbone SE module. Experimental results show that the DNE plays an important role in the SE module by increasing the quality and the intelligibility of corrupted signal even if the noise is non-stationary and unseen in training. In addition, we demonstrate that the DNE can be flexibly applied to other neural network-based SE modules.

eess.AS↗