arXiv subjects

Hong-Goo Kang

Publications and source records attributed to Hong-Goo Kang.

At least 19 recordsLinked to original sources

arXiv2026

MoLGE: Mixture of Language Group Experts for Efficient Scaling of Massively Multilingual Speech Recognition

Massively multilingual automatic speech recognition (ASR) models covering hundreds of languages must maintain robust performance across diverse linguistic and acoustic conditions. However, these models often encounter the curse of multilinguality, where model capacity is diluted across languages. To address this challenge, we propose Mixture of Language Group Experts (MoLGE), built upon speech self-supervised models (S3Ms). MoLGE assigns dedicated expert modules to clusters of similar languages, reducing the number of required submodules compared to conventional language-specific Mixture-of-Experts (MoE) schemes. It further integrates a hierarchical Low-Rank Adaptation (LoRA) strategy into the disentangled acoustic and linguistic components of the S3M architecture, enabling efficient modeling of language-specific characteristics while maintaining parameter efficiency. Further, we investigate the impact of language grouping strategies based on both linguistic and data-driven criteria on overall performance, providing an interpretable perspective on how language structure influences scalability in multilingual speech systems. In experiments, we evaluate MoLGE on a multilingual benchmark encompassing 495 languages. Results demonstrate that MoLGE consistently outperforms dense multilingual baselines with a minimal increase in trainable parameters. Notably, these language grouping strategies yield substantial improvements for both phonetic and orthographic aspects of ASR modeling. Our findings suggest that structured language specialization provides an effective pathway for massively scaling language coverage of multilingual ASR.

cs.CL↗

arXiv2026

UR-BERT: Scaling Text Encoders for Massively Multilingual TTS Through Universal Romanization and Speech Token Prediction

We propose UR-BERT, a Romanized transcription-based text-to-speech (TTS) encoder for massively multilingual TTS systems. Conventional grapheme-to-phoneme (G2P)-based approaches are limited to around 100 languages due to the availability of reliable G2P resources. In contrast, UR-BERT scales to 495 languages by unifying diverse writing systems into a shared Romanization representation. To further enhance phonetic fidelity and text-speech alignment, we introduce a speech token prediction objective during training, which encourages the encoder to learn speech-aware phonetic representations in a data-efficient manner. Experiments show that TTS systems built on UR-BERT consistently outperform recent text encoder baselines across a wide range of languages and resource conditions, and demonstrate strong generalization to unseen languages.

cs.CL↗

arXiv2026

UniverSR: Unified and Versatile Audio Super-Resolution via Vocoder-Free Flow Matching

In this paper, we present a vocoder-free framework for audio super-resolution that employs a flow matching generative model to capture the conditional distribution of complex-valued spectral coefficients. Unlike conventional two-stage diffusion-based approaches that predict a mel-spectrogram and then rely on a pre-trained neural vocoder to synthesize waveforms, our method directly reconstructs waveforms via the inverse Short-Time Fourier Transform (iSTFT), thereby eliminating the dependence on a separate vocoder. This design not only simplifies end-to-end optimization but also overcomes a critical bottleneck of two-stage pipelines, where the final audio quality is fundamentally constrained by vocoder performance. Experiments show that our model consistently produces high-fidelity 48 kHz audio across diverse upsampling factors, achieving state-of-the-art performance on both speech and general audio datasets.

eess.AS↗

arXiv2025

LAMA-UT: Language Agnostic Multilingual ASR through Orthography Unification and Language-Specific Transliteration

Building a universal multilingual automatic speech recognition (ASR) model that performs equitably across languages has long been a challenge due to its inherent difficulties. To address this task we introduce a Language-Agnostic Multilingual ASR pipeline through orthography Unification and language-specific Transliteration (LAMA-UT). LAMA-UT operates without any language-specific modules while matching the performance of state-of-the-art models trained on a minimal amount of data. Our pipeline consists of two key steps. First, we utilize a universal transcription generator to unify orthographic features into Romanized form and capture common phonetic characteristics across diverse languages. Second, we utilize a universal converter to transform these universal transcriptions into language-specific ones. In experiments, we demonstrate the effectiveness of our proposed method leveraging universal transcriptions for massively multilingual ASR. Our pipeline achieves a relative error reduction rate of 45% when compared to Whisper and performs comparably to MMS, despite being trained on only 0.1% of Whisper's training data. Furthermore, our pipeline does not rely on any language-specific modules. However, it performs on par with zero-shot ASR approaches which utilize additional language-specific lexicons and language models. We expect this framework to serve as a cornerstone for flexible multilingual ASR systems that are generalizable even to unseen languages.

cs.CL↗

arXiv2025

SAGE-LD: Towards Scalable and Generalizable End-to-End Language Diarization via Simulated Data Augmentation

In this paper, we present a neural spoken language diarization model that supports an unconstrained span of languages within a single framework. Our approach integrates a learnable query-based architecture grounded in multilingual awareness, with large-scale pretraining on simulated code-switching data. By jointly leveraging these two components, our method overcomes the limitations of conventional approaches in data scarcity and architecture optimization, and generalizes effectively to real-world multilingual settings across diverse environments. Experimental results demonstrate that our approach achieves state-of-the-art performance on several language diarization benchmarks, with a relative performance improvement of 23% to 52% over previous methods. We believe that this work not only advances research in language diarization but also establishes a foundational framework for code-switching speech technologies.

cs.CL↗

arXiv2025

MF-PAM: Accurate Pitch Estimation through Periodicity Analysis and Multi-level Feature Fusion

We introduce Multi-level feature Fusion-based Periodicity Analysis Model (MF-PAM), a novel deep learning-based pitch estimation model that accurately estimates pitch trajectory in noisy and reverberant acoustic environments. Our model leverages the periodic characteristics of audio signals and involves two key steps: extracting pitch periodicity using periodic non-periodic convolution (PNP-Conv) blocks and estimating pitch by aggregating multi-level features using a modified bi-directional feature pyramid network (BiFPN). We evaluate our model on speech and music datasets and achieve superior pitch estimation performance compared to state-of-the-art baselines while using fewer model parameters. Our model achieves 99.20 % accuracy in pitch estimation on a clean musical dataset. Overall, our proposed model provides a promising solution for accurate pitch estimation in challenging acoustic environments and has potential applications in audio signal processing.

eess.AS↗

arXiv2025

UniCoM: A Universal Code-Switching Speech Generator

Code-switching (CS), the alternation between two or more languages within a single speaker's utterances, is common in real-world conversations and poses significant challenges for multilingual speech technology. However, systems capable of handling this phenomenon remain underexplored, primarily due to the scarcity of suitable datasets. To resolve this issue, we propose Universal Code-Mixer (UniCoM), a novel pipeline for generating high-quality, natural CS samples without altering sentence semantics. Our approach utilizes an algorithm we call Substituting WORDs with Synonyms (SWORDS), which generates CS speech by replacing selected words with their translations while considering their parts of speech. Using UniCoM, we construct Code-Switching FLEURS (CS-FLEURS), a multilingual CS corpus designed for automatic speech recognition (ASR) and speech-to-text translation (S2TT). Experimental results show that CS-FLEURS achieves high intelligibility and naturalness, performing comparably to existing datasets on both objective and subjective metrics. We expect our approach to advance CS speech technology and enable more inclusive multilingual systems.

cs.CL↗

arXiv2025

Neural Spectral Band Generation for Audio Coding

Spectral band replication (SBR) enables bit-efficient coding by generating high-frequency bands from the low-frequency ones. However, it only utilizes coarse spectral features upon a subband-wise signal replication, limiting adaptability to diverse acoustic signals. In this paper, we explore the efficacy of a deep neural network (DNN)-based generative approach for coding the high-frequency bands, which we call neural spectral band generation (n-SBG). Specifically, we propose a DNN-based encoder-decoder structure to extract and quantize the side information related to the high-frequency components and generate the components given both the side information and the decoded core-band signals. The whole coding pipeline is optimized with generative adversarial criteria to enable the generation of perceptually plausible sound. From experiments using AAC as the core codec, we show that the proposed method achieves a better perceptual quality than HE-AAC-v1 with much less side information.

eess.AS↗

arXiv2025

StableQuant: Layer Adaptive Post-Training Quantization for Speech Foundation Models

In this paper, we propose StableQuant, a novel adaptive post-training quantization (PTQ) algorithm for widely used speech foundation models (SFMs). While PTQ has been successfully employed for compressing large language models (LLMs) due to its ability to bypass additional fine-tuning, directly applying these techniques to SFMs may not yield optimal results, as SFMs utilize distinct network architecture for feature extraction. StableQuant demonstrates optimal quantization performance regardless of the network architecture type, as it adaptively determines the quantization range for each layer by analyzing both the scale distributions and overall performance. We evaluate our algorithm on two SFMs, HuBERT and wav2vec2.0, for an automatic speech recognition (ASR) task, and achieve superior performance compared to traditional PTQ methods. StableQuant successfully reduces the sizes of SFM models to a quarter and doubles the inference speed while limiting the word error rate (WER) performance drop to less than 0.3% with 8-bit quantization.

eess.AS↗

arXiv2024

Optimization of DNN-based speaker verification model through efficient quantization technique

As Deep Neural Networks (DNNs) rapidly advance in various fields, including speech verification, they typically involve high computational costs and substantial memory consumption, which can be challenging to manage on mobile systems. Quantization of deep models offers a means to reduce both computational and memory expenses. Our research proposes an optimization framework for the quantization of the speaker verification model. By analyzing performance changes and model size reductions in each layer of a pre-trained speaker verification model, we have effectively minimized performance degradation while significantly reducing the model size. Our quantization algorithm is the first attempt to maintain the performance of the state-of-the-art pre-trained speaker verification model, ECAPATDNN, while significantly compressing its model size. Overall, our quantization approach resulted in reducing the model size by half, with an increase in EER limited to 0.07%.

eess.AS↗

arXiv2024

Speaker-Independent Acoustic-to-Articulatory Inversion through Multi-Channel Attention Discriminator

We present a novel speaker-independent acoustic-to-articulatory inversion (AAI) model, overcoming the limitations observed in conventional AAI models that rely on acoustic features derived from restricted datasets. To address these challenges, we leverage representations from a pre-trained self-supervised learning (SSL) model to more effectively estimate the global, local, and kinematic pattern information in Electromagnetic Articulography (EMA) signals during the AAI process. We train our model using an adversarial approach and introduce an attention-based Multi-duration phoneme discriminator (MDPD) designed to fully capture the intricate relationship among multi-channel articulatory signals. Our method achieves a Pearson correlation coefficient of 0.847, marking state-of-the-art performance in speaker-independent AAI models. The implementation details and code can be found online.

eess.SP↗

arXiv2024

Speak in the Scene: Diffusion-based Acoustic Scene Transfer toward Immersive Speech Generation

This paper introduces a novel task in generative speech processing, Acoustic Scene Transfer (AST), which aims to transfer acoustic scenes of speech signals to diverse environments. AST promises an immersive experience in speech perception by adapting the acoustic scene behind speech signals to desired environments. We propose AST-LDM for the AST task, which generates speech signals accompanied by the target acoustic scene of the reference prompt. Specifically, AST-LDM is a latent diffusion model conditioned by CLAP embeddings that describe target acoustic scenes in either audio or text modalities. The contributions of this paper include introducing the AST task and implementing its baseline model. For AST-LDM, we emphasize its core framework, which is to preserve the input speech and generate audio consistently with both the given speech and the target acoustic environment. Experiments, including objective and subjective tests, validate the feasibility and efficacy of our approach.

eess.AS↗

arXiv2024

Enhanced Deep Speech Separation in Clustered Ad Hoc Distributed Microphone Environments

Ad-hoc distributed microphone environments, where microphone locations and numbers are unpredictable, present a challenge to traditional deep learning models, which typically require fixed architectures. To tailor deep learning models to accommodate arbitrary array configurations, the Transform-Average-Concatenate (TAC) layer was previously introduced. In this work, we integrate TAC layers with dual-path transformers for speech separation from two simultaneous talkers in realistic settings. However, the distributed nature makes it hard to fuse information across microphones efficiently. Therefore, we explore the efficacy of blindly clustering microphones around sources of interest prior to enhancement. Experimental results show that this deep cluster-informed approach significantly improves the system's capacity to cope with the inherent variability observed in ad-hoc distributed microphone environments.

eess.AS↗

arXiv2023

BrainTalker: Low-Resource Brain-to-Speech Synthesis with Transfer Learning using Wav2Vec 2.0

Decoding spoken speech from neural activity in the brain is a fast-emerging research topic, as it could enable communication for people who have difficulties with producing audible speech. For this task, electrocorticography (ECoG) is a common method for recording brain activity with high temporal resolution and high spatial precision. However, due to the risky surgical procedure required for obtaining ECoG recordings, relatively little of this data has been collected, and the amount is insufficient to train a neural network-based Brain-to-Speech (BTS) system. To address this problem, we propose BrainTalker-a novel BTS framework that generates intelligible spoken speech from ECoG signals under extremely low-resource scenarios. We apply a transfer learning approach utilizing a pre-trained self supervised model, Wav2Vec 2.0. Specifically, we train an encoder module to map ECoG signals to latent embeddings that match Wav2Vec 2.0 representations of the corresponding spoken speech. These embeddings are then transformed into mel-spectrograms using stacked convolutional and transformer-based layers, which are fed into a neural vocoder to synthesize speech waveform. Experimental results demonstrate our proposed framework achieves outstanding performance in terms of subjective and objective metrics, including a Pearson correlation coefficient of 0.9 between generated and ground truth mel spectrograms. We share publicly available Demos and Code.

eess.AS↗

arXiv2023

Style Modeling for Multi-Speaker Articulation-to-Speech

In this paper, we propose a neural articulation-to-speech (ATS) framework that synthesizes high-quality speech from articulatory signal in a multi-speaker situation. Most conventional ATS approaches only focus on modeling contextual information of speech from a single speaker's articulatory features. To explicitly represent each speaker's speaking style as well as the contextual information, our proposed model estimates style embeddings, guided from the essential speech style attributes such as pitch and energy. We adopt convolutional layers and transformer-based attention layers for our model to fully utilize both local and global information of articulatory signals, measured by electromagnetic articulography (EMA). Our model significantly improves the quality of synthesized speech compared to the baseline in terms of objective and subjective measurements in the Haskins dataset.

eess.AS↗

arXiv2023

Self-supervised Complex Network for Machine Sound Anomaly Detection

In this paper, we propose an anomaly detection algorithm for machine sounds with a deep complex network trained by self-supervision. Using the fact that phase continuity information is crucial for detecting abnormalities in time-series signals, our proposed algorithm utilizes the complex spectrum as an input and performs complex number arithmetic throughout the entire process. Since the usefulness of phase information can vary depending on the type of machine sound, we also apply an attention mechanism to control the weights of the complex and magnitude spectrum bottleneck features depending on the machine type. We train our network to perform a self-supervised task that classifies the machine identifier (id) of normal input sounds among multiple classes. At test time, an input signal is detected as anomalous if the trained model is unable to correctly classify the id. In other words, we determine the presence of an anomality when the output cross-entropy score of the multiclass identification task is lower than a pre-defined threshold. Experiments with the MIMII dataset show that the proposed algorithm has a much higher area under the curve (AUC) score than conventional magnitude spectrum-based algorithms.

eess.AS↗

arXiv2023

C2C: Cough to COVID-19 Detection in BHI 2023 Data Challenge

This report describes our submission to BHI 2023 Data Competition: Sensor challenge. Our Audio Alchemists team designed an acoustic-based COVID-19 diagnosis system, Cough to COVID-19 (C2C), and won the 1st place in the challenge. C2C involves three key contributions: pre-processing of input signals, cough-related representation extraction leveraging Wav2vec2.0, and data augmentation. Through experimental findings, we demonstrate C2C's promising potential to enhance the diagnostic accuracy of COVID-19 via cough signals. Our proposed model achieves a ROC-AUC value of 0.7810 in the context of COVID-19 diagnosis. The implementation details and the python code can be found in the following link: https://github.com/Woo-jin-Chung/BHI_2023_challenge_Audio_Alchemists

eess.AS↗

arXiv2023

Pruning Self-Attention for Zero-Shot Multi-Speaker Text-to-Speech

For personalized speech generation, a neural text-to-speech (TTS) model must be successfully implemented with limited data from a target speaker. To this end, the baseline TTS model needs to be amply generalized to out-of-domain data (i.e., target speaker's speech). However, approaches to address this out-of-domain generalization problem in TTS have yet to be thoroughly studied. In this work, we propose an effective pruning method for a transformer known as sparse attention, to improve the TTS model's generalization abilities. In particular, we prune off redundant connections from self-attention layers whose attention weights are below the threshold. To flexibly determine the pruning strength for searching optimal degree of generalization, we also propose a new differentiable pruning method that allows the model to automatically learn the thresholds. Evaluations on zero-shot multi-speaker TTS verify the effectiveness of our method in terms of voice quality and speaker similarity.

cs.SD↗