1 January 2025 in Nonlinear Theory and Its Applications, IEICE
In the present paper, we investigate a novel personal authentication method using unconscious personal features of a swipe motion in the pattern lock authentication on Android devices. As unconscious personal features, we employ a velocity variation of the finger movement and the contact area of the finger on the screen during the swiping motion. In order to show the latent individual characteristics in these features, we have conducted t-tests and the result shows that about 98 % of the test points have significant differences. This indicates that the extracted features are unconscious features latent in the swiping operation. In our personal authentication method based on unconscious features in swipe motion, we have applied standardized Euclidean distance. In authentication experiments, we determine appropriate threshold values for the distance from training and validation data, we evaluate the false rejection rate (FRR) and the false acceptance rate (FAR). The best cases of FRR and FAR are 0 % and 0 %, and the worst cases are 4 % and 0 %, respectively. In smartphone personal authentication, it is crucial to minimize the FAR as much as possible, making our authentication method practical for personal authentication on smartphones.
1 January 2025 in Nonlinear Theory and Its Applications, IEICE
This study introduces a method for forecasting electricity usage outliers employing the LightGBM algorithm. Previous research has demonstrated the capability of deep learning models to predict electricity usage trends; however, these models often exhibit low accuracy when forecasting outliers that cause electricity use to jump. Accurately predicting when electricity usage will reach outlier values several hours in advance is crucial for power companies to ensure appropriate electricity supply. We employed LightGBM, a machine learning technique, to predict outliers and successfully demonstrated its effectiveness in achieving reliable prediction accuracy. Additionally, we have successfully identified the key features essential for accurate outlier prediction.
1 January 2026 in Nonlinear Theory and Its Applications, IEICE
1 January 2026 in Nonlinear Theory and Its Applications, IEICE
1 January 2026 in Nonlinear Theory and Its Applications, IEICE
1 January 2025 in Nonlinear Theory and Its Applications, IEICE
In this study, we propose a method to construct an image recognition model with a single-channel input using transfer learning and data augmentation for music emotion classification. The data augmentation method generates a variety of spectrogram images by varying the STFT window size in small increments. This method ensures data equivalent to five times the amount of the original data and prevents degradation of classification performance due to insufficient data. The model construction method using transfer learning for grayscale images is designed to adapt the pre-trained EfficientNetV2 model, which was originally trained on ImageNet. The constructed model through transfer learning and our proposed data augmentation method achieved a classification accuracy of 94.8% on the 4Q Audio Emotion Dataset. Thus, our construction method using transfer learning for grayscale images, combined with the proposed data augmentation method, is effective in achieving music high-accuracy emotion classification.