Comparative Study of Different Algorithms for Human Motion Direction Prediction Based on Multimodal Data.

Area of Science:

• Biomechanics
• Artificial Intelligence
• Wearable Technology

Background:

• Accurate human movement prediction is vital for rehabilitation, sports science, and intelligent systems.
• Existing methods may not fully capture complex spatiotemporal dynamics of human motion.

Purpose of the Study:

• To develop and evaluate a hybrid deep learning model for precise human movement direction prediction.
• To compare the performance of a CNN-BiLSTM model against other deep learning architectures.

Main Methods:

• Developed a hybrid deep learning model integrating Convolutional Neural Network (CNN) and Bidirectional Long Short-Term Memory (BiLSTM).
• Utilized plantar pressure and inertial sensor data for spatiotemporal feature learning.
• Conducted comparative experiments with CNN, BiLSTM, CNN-LSTM, and CNN-BiLSTM models.

Main Results:

• The CNN-BiLSTM model demonstrated superior performance with the lowest RMSE (0.26) and MAE (0.14), and an R² of 0.86.
• Achieved high fitting accuracy and generalization ability on the test set.
• Effectively captured local spatial features and bidirectional temporal dependencies.

Conclusions:

• The CNN-BiLSTM model offers a reliable framework for real-time human motion prediction.
• Demonstrates strong adaptability for complex motion scenarios.
• Potential applications in intelligent gait analysis, wearable monitoring, and human-machine interaction.