Institution profile

Tokyo University of Agriculture and Technology

Academic institutionasia · jp
Official website
Research library57linked papers
Opportunities0open roles
Selected work

Representative Papers

Prox-Friendly Log-Magnitude Prior on Complex-Valued Signal

Sep 28, 2026

This study addresses the challenge of incorporating log-magnitude domain priors into standard proximal splitting algorithms. To overcome this limitation, the authors propose EPILOG, an exponential penalty regularizer that innovatively introduces auxiliary variables to establish an analytical connection with the log-magnitude, thereby indirectly enforcing signal priors. Furthermore, a closed-form proximal operator is derived to construct an efficient solver for complex-valued signals. The primary contribution of this work lies in enabling a proximal-friendly treatment of log-domain priors. Experimental results on speech dereverberation validate the effectiveness of the proposed approach, demonstrating significant improvements in cepstral domain sparsity and offering a novel paradigm for complex-valued signal processing.

0 citationsRead paper

LipSSM: Structurally Lipschitz-Bounded Cascaded State-Space Model via Metric Transfer between Consecutive SSM Layers

Sep 25, 2026

This study addresses the limitations of conventional layer-wise Lipschitz constraints, which yield overly loose bound estimates and restrict the expressivity and robustness of neural networks. To overcome this, we introduce the LipKernel concept into cascaded state space models (SSMs) for the first time, constructing Lipschitz-continuous deep neural networks with rigorous theoretical guarantees via cross-layer metric propagation. By leveraging a structured Lipschitz bound theorem to facilitate inter-layer information transfer, our approach yields a substantially tighter global Lipschitz bound while effectively modeling long-range dependencies. Consequently, this work transcends conservative constraints without compromising strict robustness assurances, significantly enhancing both model expressivity and empirical performance. The proposed method is theoretically sound and practically effective.

0 citationsRead paper

TinyCardioUNet: IMU-to-ECG Translation with Graph-Encoded Inter-Axis Dependencies and Tensor Decomposition-Based Parameter Reduction

Sep 24, 2026

This study addresses the technical challenge of reconstructing electrocardiogram (ECG) signals from chest-mounted inertial measurement unit (IMU) data for electrode-free continuous heart rate monitoring. We propose a lightweight UNet model integrating graph neural networks (GNNs) and tensor decomposition. The method directly utilizes full six-axis IMU data without channel selection, employing GNNs to encode inter-axis dependencies and variational Bayesian inference to achieve automatic rank selection in tensor decomposition for substantial parameter compression. With only 36,000 parameters, the proposed architecture attains a reconstruction accuracy of 0.098 RMSE and a correlation coefficient of 0.677, while maintaining robustness under noisy conditions. These results validate the effectiveness of ultra-compact architectures for physiological signal estimation.

0 citationsRead paper
Recent publications

Latest Papers

Prox-Friendly Log-Magnitude Prior on Complex-Valued Signal

Sep 28, 2026

This study addresses the challenge of incorporating log-magnitude domain priors into standard proximal splitting algorithms. To overcome this limitation, the authors propose EPILOG, an exponential penalty regularizer that innovatively introduces auxiliary variables to establish an analytical connection with the log-magnitude, thereby indirectly enforcing signal priors. Furthermore, a closed-form proximal operator is derived to construct an efficient solver for complex-valued signals. The primary contribution of this work lies in enabling a proximal-friendly treatment of log-domain priors. Experimental results on speech dereverberation validate the effectiveness of the proposed approach, demonstrating significant improvements in cepstral domain sparsity and offering a novel paradigm for complex-valued signal processing.

0 citationsRead paper

LipSSM: Structurally Lipschitz-Bounded Cascaded State-Space Model via Metric Transfer between Consecutive SSM Layers

Sep 25, 2026

This study addresses the limitations of conventional layer-wise Lipschitz constraints, which yield overly loose bound estimates and restrict the expressivity and robustness of neural networks. To overcome this, we introduce the LipKernel concept into cascaded state space models (SSMs) for the first time, constructing Lipschitz-continuous deep neural networks with rigorous theoretical guarantees via cross-layer metric propagation. By leveraging a structured Lipschitz bound theorem to facilitate inter-layer information transfer, our approach yields a substantially tighter global Lipschitz bound while effectively modeling long-range dependencies. Consequently, this work transcends conservative constraints without compromising strict robustness assurances, significantly enhancing both model expressivity and empirical performance. The proposed method is theoretically sound and practically effective.

0 citationsRead paper

TinyCardioUNet: IMU-to-ECG Translation with Graph-Encoded Inter-Axis Dependencies and Tensor Decomposition-Based Parameter Reduction

Sep 24, 2026

This study addresses the technical challenge of reconstructing electrocardiogram (ECG) signals from chest-mounted inertial measurement unit (IMU) data for electrode-free continuous heart rate monitoring. We propose a lightweight UNet model integrating graph neural networks (GNNs) and tensor decomposition. The method directly utilizes full six-axis IMU data without channel selection, employing GNNs to encode inter-axis dependencies and variational Bayesian inference to achieve automatic rank selection in tensor decomposition for substantial parameter compression. With only 36,000 parameters, the proposed architecture attains a reconstruction accuracy of 0.098 RMSE and a correlation coefficient of 0.677, while maintaining robustness under noisy conditions. These results validate the effectiveness of ultra-compact architectures for physiological signal estimation.

0 citationsRead paper