arXiv subjects

Hyeongju Kim

Publications and source records attributed to Hyeongju Kim.

18 recordsLinked to original sources

arXiv2026

RobustSpeechFlow: Learning Robust Text-to-Speech Trajectories via Augmentation-based Contrastive Flow Matching

While flow-matching text-to-speech (TTS) achieves strong zero-shot speaker similarity and naturalness, it remains susceptible to content fidelity issues, particularly skip and repeat errors from imperfect alignment. We propose RobustSpeechFlow, a training strategy that improves alignment robustness by extending contrastive flow matching with length-preserving repeat and skip latent augmentations. Requiring no external aligners or preference data, our method directly penalizes realistic failure modes and readily integrates into existing pipelines. On Seed-TTS-eval, it reduces the word error rate (WER) from 1.44 to 1.38 using only 0.06B parameters. On our ZERO500 benchmark, it delivers consistent intelligibility improvements across diverse speaker and prosody conditions; at NFE=24, it reduces English character error rate (CER) from 0.48\% to 0.35\% and Korean CER from 0.81\% to 0.57\%. Audio samples: https://robustspeechflow.github.io/

cs.SD↗

arXiv2026

Super Monotonic Alignment Search

Monotonic alignment search (MAS), introduced by Glow-TTS, is one of the most popular algorithm in text-to-speech to estimate unknown alignments between text and speech. Since this algorithm needs to search for the most probable alignment with dynamic programming by caching all possible paths, the time complexity of the algorithm is $O(T \times S)$, where $T$ is the length of text and $S$ is the length of speech representation. The authors of Glow-TTS run this algorithm on CPU, and while they mentioned it is difficult to parallelize, we found that MAS can be parallelized in text length dimension and CPU execution consumes an inordinate amount of time for inter-device copy. Therefore, we implemented a Triton kernel and PyTorch JIT script to accelerate MAS on GPU without inter-device copy. As a result, Super-MAS Triton kernel is up to 72 times faster in the extreme-length case. The code is available at https://github.com/supertone-inc/super-monotonic-align.

eess.AS↗

arXiv2025

Robust TTS Training via Self-Purifying Flow Matching for the WildSpoof 2026 TTS Track

This paper presents a lightweight text-to-speech (TTS) system developed for the WildSpoof Challenge TTS Track. Our approach fine-tunes the recently released open-weight TTS model, \textit{Supertonic}\footnote{\url{https://github.com/supertone-inc/supertonic}}, with Self-Purifying Flow Matching (SPFM) to enable robust adaptation to in-the-wild speech. SPFM mitigates label noise by comparing conditional and unconditional flow matching losses on each sample, routing suspicious text--speech pairs to unconditional training while still leveraging their acoustic information. The resulting model achieves the lowest Word Error Rate (WER) among all participating teams, while ranking second in perceptual metrics such as UTMOS and DNSMOS. These findings demonstrate that efficient, open-weight architectures like Supertonic can be effectively adapted to diverse real-world speech conditions when combined with explicit noise-handling mechanisms such as SPFM.

cs.SD↗

arXiv2025

SupertonicTTS: Towards Highly Efficient and Streamlined Text-to-Speech System

We introduce SupertonicTTS, a novel text-to-speech (TTS) system designed for efficient and streamlined speech synthesis. SupertonicTTS comprises three components: a speech autoencoder for continuous latent representation, a text-to-latent module leveraging flow-matching for text-to-latent mapping, and an utterance-level duration predictor. To enable a lightweight architecture, we employ a low-dimensional latent space, temporal compression of latents, and ConvNeXt blocks. The TTS pipeline is further simplified by operating directly on raw character-level text and employing cross-attention for text-speech alignment, thus eliminating the need for grapheme-to-phoneme (G2P) modules and external aligners. In addition, we propose context-sharing batch expansion that accelerates loss convergence and stabilizes text-speech alignment with minimal memory and I/O overhead. Experimental results demonstrate that SupertonicTTS delivers performance comparable to contemporary zero-shot TTS models with only 44M parameters, while significantly reducing architectural complexity and computational cost. Audio samples are available at: https://supertonictts.github.io/.

eess.AS↗

arXiv2025

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.

eess.AS↗

arXiv2025

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.

eess.AS↗

arXiv2025

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.

eess.AS↗

arXiv2024

DualSpeech: Enhancing Speaker-Fidelity and Text-Intelligibility Through Dual Classifier-Free Guidance

Text-to-Speech (TTS) models have advanced significantly, aiming to accurately replicate human speech's diversity, including unique speaker identities and linguistic nuances. Despite these advancements, achieving an optimal balance between speaker-fidelity and text-intelligibility remains a challenge, particularly when diverse control demands are considered. Addressing this, we introduce DualSpeech, a TTS model that integrates phoneme-level latent diffusion with dual classifier-free guidance. This approach enables exceptional control over speaker-fidelity and text-intelligibility. Experimental results demonstrate that by utilizing the sophisticated control, DualSpeech surpasses existing state-of-the-art TTS models in performance. Demos are available at https://bit.ly/48Ewoib.

eess.AS↗

arXiv2022

Towards trustworthy phoneme boundary detection with autoregressive model and improved evaluation metric

Phoneme boundary detection has been studied due to its central role in various speech applications. In this work, we point out that this task needs to be addressed not only by algorithmic way, but also by evaluation metric. To this end, we first propose a state-of-the-art phoneme boundary detector that operates in an autoregressive manner, dubbed SuperSeg. Experiments on the TIMIT and Buckeye corpora demonstrates that SuperSeg identifies phoneme boundaries with significant margin compared to existing models. Furthermore, we note that there is a limitation on the popular evaluation metric, R-value, and propose new evaluation metrics that prevent each boundary from contributing to evaluation multiple times. The proposed metrics reveal the weaknesses of non-autoregressive baselines and establishes a reliable criterion that suits for evaluating phoneme boundary detection.

cs.SD↗

arXiv2022

NANSY++: Unified Voice Synthesis with Neural Analysis and Synthesis

Various applications of voice synthesis have been developed independently despite the fact that they generate "voice" as output in common. In addition, most of the voice synthesis models still require a large number of audio data paired with annotated labels (e.g., text transcription and music score) for training. To this end, we propose a unified framework of synthesizing and manipulating voice signals from analysis features, dubbed NANSY++. The backbone network of NANSY++ is trained in a self-supervised manner that does not require any annotations paired with audio. After training the backbone network, we efficiently tackle four voice applications - i.e. voice conversion, text-to-speech, singing voice synthesis, and voice designing - by partially modeling the analysis features required for each task. Extensive experiments show that the proposed framework offers competitive advantages such as controllability, data efficiency, and fast training convergence, while providing high quality synthesis. Audio samples: tinyurl.com/8tnsy3uc.

cs.SD↗

arXiv2022

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.

cs.SD↗

arXiv2021

MLP Singer: Towards Rapid Parallel Korean Singing Voice Synthesis

Recent developments in deep learning have significantly improved the quality of synthesized singing voice audio. However, prominent neural singing voice synthesis systems suffer from slow inference speed due to their autoregressive design. Inspired by MLP-Mixer, a novel architecture introduced in the vision literature for attention-free image classification, we propose MLP Singer, a parallel Korean singing voice synthesis system. To the best of our knowledge, this is the first work that uses an entirely MLP-based architecture for voice synthesis. Listening tests demonstrate that MLP Singer outperforms a larger autoregressive GAN-based system, both in terms of audio quality and synthesis speed. In particular, MLP Singer achieves a real-time factor of up to 200 and 3400 on CPUs and GPUs respectively, enabling order of magnitude faster generation on both environments.

cs.SD↗

arXiv2021

Diff-TTS: A Denoising Diffusion Model for Text-to-Speech

Although neural text-to-speech (TTS) models have attracted a lot of attention and succeeded in generating human-like speech, there is still room for improvements to its naturalness and architectural efficiency. In this work, we propose a novel non-autoregressive TTS model, namely Diff-TTS, which achieves highly natural and efficient speech synthesis. Given the text, Diff-TTS exploits a denoising diffusion framework to transform the noise signal into a mel-spectrogram via diffusion time steps. In order to learn the mel-spectrogram distribution conditioned on the text, we present a likelihood-based optimization method for TTS. Furthermore, to boost up the inference speed, we leverage the accelerated sampling method that allows Diff-TTS to generate raw waveforms much faster without significantly degrading perceptual quality. Through experiments, we verified that Diff-TTS generates 28 times faster than the real-time with a single NVIDIA 2080Ti GPU.

eess.AS↗

arXiv2021

Continuous Monitoring of Blood Pressure with Evidential Regression

Photoplethysmogram (PPG) signal-based blood pressure (BP) estimation is a promising candidate for modern BP measurements, as PPG signals can be easily obtained from wearable devices in a non-invasive manner, allowing quick BP measurement. However, the performance of existing machine learning-based BP measuring methods still fall behind some BP measurement guidelines and most of them provide only point estimates of systolic blood pressure (SBP) and diastolic blood pressure (DBP). In this paper, we present a cutting-edge method which is capable of continuously monitoring BP from the PPG signal and satisfies healthcare criteria such as the Association for the Advancement of Medical Instrumentation (AAMI) and the British Hypertension Society (BHS) standards. Furthermore, the proposed method provides the reliability of the predicted BP by estimating its uncertainty to help diagnose medical condition based on the model prediction. Experiments on the MIMIC II database verify the state-of-the-art performance of the proposed method under several metrics and its ability to accurately represent uncertainty in prediction.

eess.SP↗

arXiv2021

Gated Recurrent Context: Softmax-free Attention for Online Encoder-Decoder Speech Recognition

Recently, attention-based encoder-decoder (AED) models have shown state-of-the-art performance in automatic speech recognition (ASR). As the original AED models with global attentions are not capable of online inference, various online attention schemes have been developed to reduce ASR latency for better user experience. However, a common limitation of the conventional softmax-based online attention approaches is that they introduce an additional hyperparameter related to the length of the attention window, requiring multiple trials of model training for tuning the hyperparameter. In order to deal with this problem, we propose a novel softmax-free attention method and its modified formulation for online attention, which does not need any additional hyperparameter at the training phase. Through a number of ASR experiments, we demonstrate the tradeoff between the latency and performance of the proposed online attention technique can be controlled by merely adjusting a threshold at the test phase. Furthermore, the proposed methods showed competitive performance to the conventional global and online attentions in terms of word-error-rates (WERs).

eess.AS↗

arXiv2020

SoftFlow: Probabilistic Framework for Normalizing Flow on Manifolds

Flow-based generative models are composed of invertible transformations between two random variables of the same dimension. Therefore, flow-based models cannot be adequately trained if the dimension of the data distribution does not match that of the underlying target distribution. In this paper, we propose SoftFlow, a probabilistic framework for training normalizing flows on manifolds. To sidestep the dimension mismatch problem, SoftFlow estimates a conditional distribution of the perturbed input data instead of learning the data distribution directly. We experimentally show that SoftFlow can capture the innate structure of the manifold data and generate high-quality samples unlike the conventional flow-based models. Furthermore, we apply the proposed framework to 3D point clouds to alleviate the difficulty of forming thin structures for flow-based models. The proposed model for 3D point clouds, namely SoftPointFlow, can estimate the distribution of various shapes more accurately and achieves state-of-the-art performance in point cloud generation.

cs.CV↗

arXiv2020

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.

cs.SD↗

arXiv2020

WaveNODE: A Continuous Normalizing Flow for Speech Synthesis

In recent years, various flow-based generative models have been proposed to generate high-fidelity waveforms in real-time. However, these models require either a well-trained teacher network or a number of flow steps making them memory-inefficient. In this paper, we propose a novel generative model called WaveNODE which exploits a continuous normalizing flow for speech synthesis. Unlike the conventional models, WaveNODE places no constraint on the function used for flow operation, thus allowing the usage of more flexible and complex functions. Moreover, WaveNODE can be optimized to maximize the likelihood without requiring any teacher network or auxiliary loss terms. We experimentally show that WaveNODE achieves comparable performance with fewer parameters compared to the conventional flow-based vocoders.

cs.SD↗