Yukikazu Nakamoto

Dr. Yukikazu Nakamoto


Research Focus & Metrics

Academic Career Length

9Years (2018-Present)

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

Representative Publications

View All Publications
Verified
Journal Article
Open access
PDF file
TZTianyu Zhang
Tianyu Zhang
QZQian Zhao
Qian Zhao
KSKilho Shin
Kilho Shin
YNYukikazu Nakamoto
Yukikazu Nakamoto
We propose a new peak searching algorithm (PSA) that uses Bayesian optimization to find probability peaks in a dataset, thereby increasing the speed and accuracy of clustering algorithms. Wireless sensor networks (WSNs) are becoming increasingly common in a wide variety of applications that analyze and use collected sensing data. Typically, the collected data cannot be directly used in modern data analysis problems that adopt machine learning techniques because such data lacks additional information (such as data labels) specifying its purpose of users. Clustering algorithms that divide the data in a dataset into clusters are often used when additional information is not provided. However, traditional clustering algorithms such as expectation–maximization (EM) and k - m e a n s algorithms require massive numbers of iterations to form clusters. Processing speeds are therefore slow, and clustering results become less accurate because of the way such algorithms form clusters. The PSA addresses these problems, and we adapt it for use with the EM and k - m e a n s algorithms, creating the modified P S E M and P S k - m e a n s algorithms. Our simulation results show that our proposed P S E M and P S k - m e a n s algorithms significantly decrease the required number of clustering iterations (by 1.99 to 6.3 times), and produce clustering that, for a synthetic dataset, is 1.69 to 1.71 times more accurate than it is for traditional EM and enhanced k - m e a n s ( k - m e a n s ++) algorithms. Moreover, in a simulation of WSN applications aimed at detecting outliers, P S E M correctly identified the outliers in a real dataset, decreasing iterations by approximately 1.88 times, and P S E M was 1.29 times more accurate than EM at a maximum.

14 February 2023 in Applied Sciences
Verified
Journal Article
Open access
PDF file
EKElhard James Kumalija
Elhard James Kumalija
YNYukikazu Nakamoto
Yukikazu Nakamoto
In IP audio systems, audio quality is degraded by environmental noise, poor network quality, and encoding–decoding algorithms. Therefore, there is a need for a continuous automatic quality evaluation of the transmitted audio. Speech quality monitoring in VoIP systems enables autonomous system adaptation. Furthermore, there are diverse IP audio transmitters and receivers, from high-performance computers and mobile phones to low-memory and low-computing-capacity embedded systems. This paper proposes MiniatureVQNet, a single-ended speech quality evaluation method for VoIP audio applications based on a lightweight deep neural network (DNN) model. The proposed model can predict the audio quality independent of the source of degradation, whether noise or network, and is light enough to run in embedded systems. Two variations of the proposed MiniatureVQNet model were evaluated: a MiniatureVQNet model trained on a dataset that contains environmental noise only, referred to as MiniatureVQNet–Noise, and a second model trained on both noise and network distortions, referred to as MiniatureVQNet–Noise–Network. The proposed MiniatureVQNet model outperforms the traditional P.563 method in terms of accuracy on all tested network conditions and environmental noise parameters. The mean squared error (MSE) of the models compared to the PESQ score for ITU-T P.563, MiniatureVQNet-Noise, and MiniatureVQNet–Noise–Network was 2.19, 0.34, and 0.21, respectively. The performance of both the MiniatureVQNet–Noise–Network and MiniatureVQNet-Noise model depends on the noise type for an SNR greater than 0 dB and less than 10 dB. In addition, training on a noise–network-distorted speech dataset improves the model prediction accuracy in all VoIP environment distortions compared to training the model on a noise-only dataset.