Mr. Pengyu Guo (Doctoral student, EEIS, Nakayama Laboratory, D3) received the Best Presentation Award at SPML 2025.

Name of award and short explanation about the award

The Best Presentation Award is presented to honor outstanding presentations delivered in each session of SPML 2025.

About awarded research(activity)

Pengyu Guo, Masaya Nakayama, CNN-Transformer-Bi-LSTM: A Hybrid Deep Learning Framework for Wearable Sensor-based Human Activity Recognition, 2025 IEEE 8th International Conference on Signal Processing and Machine Learning (SPML 2025), July, 2025. Human activity recognition (HAR) focuses on identifying and classifying human activities based on data collected from various sources. Its importance lies in its wide range of applications, especially in health monitoring, safety enhancement, efficiency, and quality of life improvement. In this work, we propose a CNN-Transformer-Bi-LSTM (CTBL) hybrid framework for wearable sensor HAR, integrating CNN for local feature extraction, Transformer encoder for long-range temporal modeling, and BiLSTM layers for bidirectional sequential learning. We evaluate our model on four benchmark datasets (MobiAct, UniMiB SHAR, UCI HAR, USC-HAD), achieving F1 scores of 91.25%, 93.14%, 89.59%, and 91.46%, respectively. Further analysis shows that the CNN layers contribute significantly to the classification of static activities, while the Bi- LSTM improves the recognition of symmetric transitional motions, highlighting the impact of different model components on HAR performance.

Your impression & future plan

I am truly honored to have received the Best Presentation Award. I would also like to express my sincere gratitude to Professor Masaya Nakayama for his guidance on this research. I will continue to pursue my research further.