Hyeongju Kim
Research Subjects
Hyeongju Kim has a total of 55 co-authors. Here are the top co-authors by number of shared publications:
- Ji-Joon Song
Korea Advanced Institute of Science and Technology
4 shared publications Juheon Lee
3 shared publications- Hans Hebert
KTH Royal Inst Technol, Sch Engn Sci Chem Biotechnol & Hlth, Dept Biomed Engn & Hlth Syst, S-14152 Huddinge, Sweden
2 shared publications Jacob Morton
Supertone Inc Seoul Republic of Korea
2 shared publicationsJinhyeok Yang
Supertone Inc Seoul Republic of Korea
2 shared publications- Maria Josefina Marcaida
Institute of Bioengineering School of Life Sciences Ecole Polytechnique Fédérale de Lausanne Lausanne Switzerland
2 shared publications
Latest Publications
- Open Access
Improving Test-Time Performance of RVQ-based Neural Codecs
The residual vector quantization (RVQ) technique plays a central role in recent advances in neural audio codecs. These models effectively synthesize high-fidelity audio from a limited number of codes due to the hierarchical structure among quantization levels. In this paper, we propose an encoding algorithm to further enhance the synthesis quality of RVQ-based neural codecs at test-time. Firstly, we point out the suboptimal nature of quantized vectors generated by conventional methods. We demonstrate that quantization error can be mitigated by selecting a different set of codes. Subsequently, we present our encoding algorithm, designed to identify a set of discrete codes that achieve a lower quantization error. We then apply the proposed method to pre-trained models and evaluate its efficacy using diverse metrics. Our experimental findings validate that our method not only reduces quantization errors, but also improves synthesis quality.
Open AccessImproving Test-Time Performance of RVQ-based Neural Codecs
The residual vector quantization (RVQ) technique plays a central role in recent advances in neural audio codecs. These models effectively synthesize high-fidelity audio from a limited number of codes due to the hierarchical structure among quantization levels. In this paper, we propose an encoding algorithm to further enhance the synthesis quality of RVQ-based neural codecs at test-time. Firstly, we point out the suboptimal nature of quantized vectors generated by conventional methods. We demonstrate that quantization error can be mitigated by selecting a different set of codes. Subsequently, we present our encoding algorithm, designed to identify a set of discrete codes that achieve a lower quantization error. We then apply the proposed method to pre-trained models and evaluate its efficacy using diverse metrics. Our experimental findings validate that our method not only reduces quantization errors, but also improves synthesis quality.
- Open Access
Training Flow Matching Models with Reliable Labels via Self-Purification
Training datasets are inherently imperfect, often containing mislabeled samples due to human annotation errors, limitations of tagging models, and other sources of noise. Such label contamination can significantly degrade the performance of a trained model. In this work, we introduce Self-Purifying Flow Matching (SPFM), a principled approach to filtering unreliable data within the flow-matching framework. SPFM identifies suspicious data using the model itself during the training process, bypassing the need for pretrained models or additional modules. Our experiments demonstrate that models trained with SPFM generate samples that accurately adhere to the specified conditioning, even when trained on noisy labels. Furthermore, we validate the robustness of SPFM on the TITW dataset, which consists of in-the-wild speech data, achieving performance that surpasses existing baselines.
Open AccessTraining Flow Matching Models with Reliable Labels via Self-Purification
Training datasets are inherently imperfect, often containing mislabeled samples due to human annotation errors, limitations of tagging models, and other sources of noise. Such label contamination can significantly degrade the performance of a trained model. In this work, we introduce Self-Purifying Flow Matching (SPFM), a principled approach to filtering unreliable data within the flow-matching framework. SPFM identifies suspicious data using the model itself during the training process, bypassing the need for pretrained models or additional modules. Our experiments demonstrate that models trained with SPFM generate samples that accurately adhere to the specified conditioning, even when trained on noisy labels. Furthermore, we validate the robustness of SPFM on the TITW dataset, which consists of in-the-wild speech data, achieving performance that surpasses existing baselines.
- Open Access
- 19 September 2025
- Published by American Association for the Advancement of Science (AAAS) in Science Advances
- Vol. 11 (38) , eadw4124
Structure of the Huntingtin F-actin complex reveals its role in cytoskeleton organization
The Huntingtin protein (HTT), named for its role in Huntington’s disease, has been best understood as a scaffolding protein that promotes vesicle transport by molecular motors along microtubules. Here, we show that HTT also interacts with the actin cytoskeleton, and its loss of function disturbs the morphology and function of the axonal growth cone. We demonstrate that HTT organizes F-actin into bundles. Cryo–electron tomography (cryo-ET) and subtomogram averaging (STA) structural analyses reveal that HTT’s N-terminal HEAT and Bridge domains wrap around F-actin, while the C-terminal HEAT domain is displaced; furthermore, HTT dimerizes via the N-HEAT domain to bridge parallel actin filaments separated by ~20 nanometers. Our study provides the structural basis for understanding how HTT interacts with and organizes the actin cytoskeleton.
- 19 September 2025
- Published by American Association for the Advancement of Science (AAAS) in Science Advances
- Vol. 11 (38) , eadw4124
Open AccessStructure of the Huntingtin F-actin complex reveals its role in cytoskeleton organization
The Huntingtin protein (HTT), named for its role in Huntington’s disease, has been best understood as a scaffolding protein that promotes vesicle transport by molecular motors along microtubules. Here, we show that HTT also interacts with the actin cytoskeleton, and its loss of function disturbs the morphology and function of the axonal growth cone. We demonstrate that HTT organizes F-actin into bundles. Cryo–electron tomography (cryo-ET) and subtomogram averaging (STA) structural analyses reveal that HTT’s N-terminal HEAT and Bridge domains wrap around F-actin, while the C-terminal HEAT domain is displaced; furthermore, HTT dimerizes via the N-HEAT domain to bridge parallel actin filaments separated by ~20 nanometers. Our study provides the structural basis for understanding how HTT interacts with and organizes the actin cytoskeleton.
- Open Access
Length-Aware Rotary Position Embedding for Text-Speech Alignment
Many recent text-to-speech (TTS) systems are built on transformer architectures and employ cross-attention mechanisms for text-speech alignment. Within these systems, rotary position embedding (RoPE) is commonly used to encode positional information in text and speech representations. In this work, we introduce length-aware RoPE (LARoPE), a simple yet effective extension of RoPE that improves text-speech alignment. Unlike RoPE, which relies on absolute indices, LARoPE computes relative distances between query and key positions using length-normalized indices. Experimental results show that LARoPE consistently outperforms RoPE, offering faster loss convergence, more accurate text-speech alignment, and higher overall TTS quality. Furthermore, LARoPE demonstrates greater resilience to variations in utterance duration and maintains stable performance in extended speech generation up to 30 seconds, whereas RoPE suffers from notable degradation. Notably, our method achieves a state-of-the-art word error rate on a standard zero-shot TTS benchmark.
Open AccessLength-Aware Rotary Position Embedding for Text-Speech Alignment
Many recent text-to-speech (TTS) systems are built on transformer architectures and employ cross-attention mechanisms for text-speech alignment. Within these systems, rotary position embedding (RoPE) is commonly used to encode positional information in text and speech representations. In this work, we introduce length-aware RoPE (LARoPE), a simple yet effective extension of RoPE that improves text-speech alignment. Unlike RoPE, which relies on absolute indices, LARoPE computes relative distances between query and key positions using length-normalized indices. Experimental results show that LARoPE consistently outperforms RoPE, offering faster loss convergence, more accurate text-speech alignment, and higher overall TTS quality. Furthermore, LARoPE demonstrates greater resilience to variations in utterance duration and maintains stable performance in extended speech generation up to 30 seconds, whereas RoPE suffers from notable degradation. Notably, our method achieves a state-of-the-art word error rate on a standard zero-shot TTS benchmark.
- Open Access
- 8 September 2022
- Published by American Society for Clinical Investigation in JCI Insight
- Vol. 7 (17)
A pathogenic proteolysis–resistant huntingtin isoform induced by an antisense oligonucleotide maintains huntingtin function
Huntington’s disease (HD) is a late-onset neurological disorder for which therapeutics are not available. Its key pathological mechanism involves the proteolysis of polyglutamine-expanded (polyQ-expanded) mutant huntingtin (mHTT), which generates N-terminal fragments containing polyQ, a key contributor to HD pathogenesis. Interestingly, a naturally occurring spliced form of HTT mRNA with truncated exon 12 encodes an HTT (HTT Δ12 ) with a deletion near the caspase-6 cleavage site. In this study, we used a multidisciplinary approach to characterize the therapeutic potential of targeting HTT exon 12. We show that HTT Δ12 was resistant to caspase-6 cleavage in both cell-free and tissue lysate assays. However, HTT Δ12 retained overall biochemical and structural properties similar to those of wt-HTT. We generated mice in which HTT exon 12 was truncated and found that the canonical exon 12 was dispensable for the main physiological functions of HTT, including embryonic development and intracellular trafficking. Finally, we pharmacologically induced HTT Δ12 using the antisense oligonucleotide (ASO) QRX-704. QRX-704 showed predictable pharmacology and efficient biodistribution. In addition, it was stable for several months and inhibited pathogenic proteolysis. Furthermore, QRX-704 treatments resulted in a reduction of HTT aggregation and an increase in dendritic spine count. Thus, ASO-induced HTT exon 12 splice switching from HTT may provide an alternative therapeutic strategy for HD.
- 8 September 2022
- Published by American Society for Clinical Investigation in JCI Insight
- Vol. 7 (17)
Open AccessA pathogenic proteolysis–resistant huntingtin isoform induced by an antisense oligonucleotide maintains huntingtin function
Huntington’s disease (HD) is a late-onset neurological disorder for which therapeutics are not available. Its key pathological mechanism involves the proteolysis of polyglutamine-expanded (polyQ-expanded) mutant huntingtin (mHTT), which generates N-terminal fragments containing polyQ, a key contributor to HD pathogenesis. Interestingly, a naturally occurring spliced form of HTT mRNA with truncated exon 12 encodes an HTT (HTT Δ12 ) with a deletion near the caspase-6 cleavage site. In this study, we used a multidisciplinary approach to characterize the therapeutic potential of targeting HTT exon 12. We show that HTT Δ12 was resistant to caspase-6 cleavage in both cell-free and tissue lysate assays. However, HTT Δ12 retained overall biochemical and structural properties similar to those of wt-HTT. We generated mice in which HTT exon 12 was truncated and found that the canonical exon 12 was dispensable for the main physiological functions of HTT, including embryonic development and intracellular trafficking. Finally, we pharmacologically induced HTT Δ12 using the antisense oligonucleotide (ASO) QRX-704. QRX-704 showed predictable pharmacology and efficient biodistribution. In addition, it was stable for several months and inhibited pathogenic proteolysis. Furthermore, QRX-704 treatments resulted in a reduction of HTT aggregation and an increase in dendritic spine count. Thus, ASO-induced HTT exon 12 splice switching from HTT may provide an alternative therapeutic strategy for HD.
- Open Access
EdiTTS: Score-based Editing for Controllable Text-to-Speech
We present EdiTTS, an off-the-shelf speech editing methodology based on score-based generative modeling for text-to-speech synthesis. EdiTTS allows for targeted, granular editing of audio, both in terms of content and pitch, without the need for any additional training, task-specific optimization, or architectural modifications to the score-based model backbone. Specifically, we apply coarse yet deliberate perturbations in the Gaussian prior space to induce desired behavior from the diffusion model while applying masks and softening kernels to ensure that iterative edits are applied only to the target region. Through listening tests and speech-to-text back transcription, we show that EdiTTS outperforms existing baselines and produces robust samples that satisfy user-imposed requirements.
Open AccessEdiTTS: Score-based Editing for Controllable Text-to-Speech
We present EdiTTS, an off-the-shelf speech editing methodology based on score-based generative modeling for text-to-speech synthesis. EdiTTS allows for targeted, granular editing of audio, both in terms of content and pitch, without the need for any additional training, task-specific optimization, or architectural modifications to the score-based model backbone. Specifically, we apply coarse yet deliberate perturbations in the Gaussian prior space to induce desired behavior from the diffusion model while applying masks and softening kernels to ensure that iterative edits are applied only to the target region. Through listening tests and speech-to-text back transcription, we show that EdiTTS outperforms existing baselines and produces robust samples that satisfy user-imposed requirements.
- Open Access
- 16 October 2021
- Published by Elsevier in STAR Protocols
- Vol. 2 (4) , 100886
Purification of full-length recombinant human huntingtin proteins with allelic series of polyglutamine lengths
- 16 October 2021
- Published by Elsevier in STAR Protocols
- Vol. 2 (4) , 100886
Open AccessPurification of full-length recombinant human huntingtin proteins with allelic series of polyglutamine lengths
- Open Access
- 1 July 2020
- Published by International Joint Conferences on Artificial Intelligence
- p. 3744-3750
Robust Front-End for Multi-Channel ASR using Flow-Based Density Estimation
For multi-channel speech recognition, speech enhancement techniques such as denoising or dereverberation are conventionally applied as a front-end processor. Deep learning-based front-ends using such techniques require aligned clean and noisy speech pairs which are generally obtained via data simulation. Recently, several joint optimization techniques have been proposed to train the front-end without parallel data within an end-to-end automatic speech recognition (ASR) scheme. However, the ASR objective is sub-optimal and insufficient for fully training the front-end, which still leaves room for improvement. In this paper, we propose a novel approach which incorporates flow-based density estimation for the robust front-end using non-parallel clean and noisy speech. Experimental results on the CHiME-4 dataset show that the proposed method outperforms the conventional techniques where the front-end is trained only with ASR objective.
- 1 July 2020
- Published by International Joint Conferences on Artificial Intelligence
- p. 3744-3750
Open AccessRobust Front-End for Multi-Channel ASR using Flow-Based Density Estimation
For multi-channel speech recognition, speech enhancement techniques such as denoising or dereverberation are conventionally applied as a front-end processor. Deep learning-based front-ends using such techniques require aligned clean and noisy speech pairs which are generally obtained via data simulation. Recently, several joint optimization techniques have been proposed to train the front-end without parallel data within an end-to-end automatic speech recognition (ASR) scheme. However, the ASR objective is sub-optimal and insufficient for fully training the front-end, which still leaves room for improvement. In this paper, we propose a novel approach which incorporates flow-based density estimation for the robust front-end using non-parallel clean and noisy speech. Experimental results on the CHiME-4 dataset show that the proposed method outperforms the conventional techniques where the front-end is trained only with ASR objective.