Multimodal Wearable Sensor Dataset for Sports Training Activity Recognition and Safety Monitoring

This dataset contains time-series sensor data collected from 45 athletes (aged 18–35) performing six typical sports training exercises: running, jumping, squatting, push-ups, sit-ups, and stretching. Each participant performed approximately 5 minutes of each activity, resulting in over 120,000 labeled samples. Data were recorded using wearable inertial measurement units (IMUs) placed on the wrist, chest, and ankle, capturing tri-axial acceleration, gyroscope, and magnetometer signals. The dataset is designed to support research in human activity recognition, sports performance analysis, and real-time safety monitoring in IoT-enabled fitness applications. It complements existing public HAR benchmarks by focusing on structured sports movements and is suitable for developing and evaluating deep learning models such as CRNNs for spatiotemporal pattern analysis.