Kai Li

Kai Li

Auto-generated profile, based on publicly available information

Kai Li

Graduate School of Advanced Science and Technology, Japan Advanced Institute of Science and Technology, Nomi 923-1292, Ishikawa, Japan

Research Focus & Metrics

Academic Career Length

2Years (2025-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
1 April 2025 in IEICE Transactions on Information and Systems
Verified
Journal Article
Open access
PDF file
MUMasashi Unoki
Masashi Unoki
KLKai Li
Kai Li
ACAnuwat Chaiwongyen
Anuwat Chaiwongyen
QNQuoc-Huy Nguyen
Quoc-Huy Nguyen
Khalid ZamanKhalid Zaman
Khalid Zaman
Skillfully fabricated artificial replicas of authentic media using advanced AI-based generators are known as “deepfakes.” Deepfakes have become a growing concern due to their increased distribution in cyber-physical spaces. In particular, deepfake speech, which is fabricated by using advanced AI-based speech analysis/synthesis techniques, can be abused for spoofing and tampering with authentic speech signals. This can enable attackers to commit serious offenses such as fraud by voice impersonation and unauthorized speaker verification. Our research project aims to construct the basis of auditory-media signal processing for defending against deepfake speech attacks. To this end, we introduce current challenges and state-of-the-art techniques for deepfake speech detection and examine current trends and remaining issues. We then introduce the basis of the acoustical features related to auditory perception and propose methods for detecting deepfake speech based on auditory-media signal processing consisting of these features and deep neural networks (DNNs).

16 May 2025 in IEEE Transactions on Audio, Speech and Language Processing
Verified
Journal Article
Open access
KLKai Li
Kai Li
Khalid ZamanKhalid Zaman
Khalid Zaman
XLXingfeng Li
Xingfeng Li
MAMasato Akagi
Masato Akagi
JDJianwu Dang
Jianwu Dang
MUMasashi Unoki
Masashi Unoki
Early detection of factory machinery malfunctions is crucial in industrial applications. In machine anomalous sound detection (ASD), different machines exhibit unique vibration-frequency ranges based on their physical properties. Meanwhile, the human auditory system is adept at tracking both temporal and spectral dynamics of machine sounds. Consequently, integrating the computational auditory models of the human auditory system with machine-specific properties can be an effective approach to machine ASD. We first quantified the frequency importances of four types of machines using the Fisher ratio (F-ratio). The quantified frequency importances were then used to design machine-specific non-uniform filterbanks (NUFBs), which extract the log non-uniform spectrum (LNS) feature. The designed NUFBs have a narrower bandwidth and higher filter distribution density in frequency regions with relatively high F-ratios. Finally, spectral and temporal modulation representations derived from the LNS feature were proposed. These proposed LNS feature and modulation representations are input into an autoencoder neural-network-based detector for ASD. The quantification results from the training set of the Malfunctioning Industrial Machine Investigation and Inspection dataset with a signal-to-noise (SNR) of 6 dB reveal that the distinguishing information between normal and anomalous sounds of different machines is encoded non-uniformly in the frequency domain. By highlighting these important frequency regions using NUFBs, the LNS feature can significantly enhance performance using the metric of AUC (area under the receiver operating characteristic curve) under various SNR conditions. Furthermore, modulation representations can further improve performance. Specifically, temporal modulation is effective for fans, pumps, and sliders, while spectral modulation is particularly effective for valves.

29 May 2025 in IEEE Transactions on Audio, Speech and Language Processing
Verified
Journal Article
Closed access
LWLijun Wang
Lijun Wang
YLYixian Lu
Yixian Lu
ZGZiyan Gao
Ziyan Gao
KLKai Li
Kai Li
JHJianqiang Huang
Jianqiang Huang
Shogo OkadaShogo Okada
Shogo Okada
Room acoustical parameters (RAPs), room geometrical parameters (RGPs) and instantaneous occupancy levels are essential metrics for parameterizing the room acoustical characteristics (RACs) of a sound field around a listener's local environment, offering comprehensive indications for various applications. Current blind estimation methods either fail to cover a broad range of real-world acoustic environments in the context of real background noise or estimate only a few RAPs and RGPs from noisy single-channel speech signals. In addition, they are limited in their ability to estimate the instantaneous occupancy level. In this paper, we propose BERP, a new universal approach to blindly estimate RAPs, RGPs, and occupancy levels. It consists of two modules: one for RAPs and RGPs and another for occupancy levels. For the former task, we use a shared room feature encoder that combines attention mechanisms with convolutional layers to learn common features across room parameters, and multiple separate parametric predictors for continuous estimation of each parameter in parallel. The combination of attention and convolutions enables the model to capture acoustic features both locally and globally from speech, yielding more robust and multitask generalizable common features. Separate predictors allow the model to independently optimize for each room parameter to reduce task learning conflict and improve per-task performance. This architecture enables universal and efficient estimation of room parameters while maintaining satisfactory performance. For occupancy level estimation, we reuse the identical encoder with a classification head, exploiting the encoder's strong sequencefeature extraction. To evaluate the effectiveness of the proposed approach, we compile a task-specific dataset from several publicly available datasets, including synthetic and real reverberant recordings. The results reveal that BERP achieves state-of-the-art (SOTA) performance and excellent adaptability to real-world scenarios. The code, datasets, and weights are available on GitHub.

1 January 2026 in IEEE Access
Verified
Journal Article
Open access
Khalid ZamanKhalid Zaman
Khalid Zaman
KLKai Li
Kai Li
ACAnuwat Chaiwongyen
Anuwat Chaiwongyen
YUYasufumi Uezu
Yasufumi Uezu
SKShunsuke Kidani
Shunsuke Kidani
MUMasashi Unoki
Masashi Unoki

20 February 2026 in Applied Sciences
Verified
Journal Article
Open access
PDF file
ACAnuwat Chaiwongyen
Anuwat Chaiwongyen
Khalid ZamanKhalid Zaman
Khalid Zaman
KLKai Li
Kai Li
SDSuradej Duangpummet
Suradej Duangpummet
JKJessada Karnjana
Jessada Karnjana
MUMasashi Unoki
Masashi Unoki
The detection of deepfake speech has become a significant research area due to rapid advancements in generative AI for speech synthesis. These technologies pose significant security risks in applications such as biometric authentication, voice-controlled systems, and automatic speaker verification (ASV) systems. Therefore, enhancing the detection capabilities of such applications is essential to mitigate potential threats. This study investigates perceptual speech-pathological features, which are commonly used to evaluate the unnaturalness of voice disorders in clinical settings, as potential indicators for detecting deepfake speech. Specifically, the timbral attributes of hardness, depth, brightness, roughness, sharpness, warmth, boominess, and reverberation are examined. The analysis reveals that these attributes provide meaningful distinctions between genuine and synthetic speech. Furthermore, the detection performance is enhanced by extending the dimensional representation of timbral attributes, enabling a more comprehensive characterization of the speech signal. This paper proposes a method that combines two models: one utilizing the different dimensions of speech-pathological features with a deep neural network (DNN), and another employing a gammatone filterbank model that simulates the auditory processing mechanism of the human cochlea with ResNet-18 architecture, improving deepfake speech detection. The proposed method is evaluated on the Automatic Speaker Verification Spoofing and Countermeasures Challenge (ASVspoof) 2019 dataset. Experimental results demonstrate that the proposed approach outperforms baseline models in terms of Equal Error Rate (EER), achieving an EER of 5.93%.

1 November 2025 in Journal of Signal Processing
Verified
Journal Article
Open access
PDF file
ISIslam J. A. M. Samiul
Islam J. A. M. Samiul
Khalid ZamanKhalid Zaman
Khalid Zaman
KLKai Li
Kai Li
ACAnuwat Chaiwongyen
Anuwat Chaiwongyen
Shogo OkadaShogo Okada
Shogo Okada
MUMasashi Unoki
Masashi Unoki
The rapid advancements in artificial intelligence (AI) technologies for singing voice synthesis have revolutionized music production but introduced significant challenges, including the misuse of AI-generated voices for deepfake purposes. This study proposes a framework for detecting deepfake singing voices using data augmentation techniques, the Synthetic Minority Over- sampling Technique (SMOTE), time stretching, and time shifting, combined with RawNet-based waveform feature extraction integrated into the Audio Anti-Spoofing Using Integrated Spectro-Temporal Graph Attention Networks (AASIST) model. The method effectively captures subtle temporal and spectral artifacts, achieving an equal error rate (EER) of 10.39% on the SVDD SingFake 2024 dataset, outperforming baseline models. This work provides a robust solution to safeguard the authenticity of AI-generated singing voices in digital media.