A Method for Predicting Trajectories of Concealed Targets via a Hybrid Decomposition and State Prediction Framework

Area of Science:

• Millimeter-wave sensing
• Signal processing
• Machine learning for target tracking

Background:

• Accurate trajectory prediction of concealed targets is difficult in complex, noisy environments for millimeter-wave (mmWave) sensors.
• Existing tracking systems struggle with environmental interference and intentional target concealment.

Purpose of the Study:

• To develop an advanced autonomous prediction framework for concealed targets using mmWave sensor data.
• To enhance the accuracy and robustness of trajectory prediction in challenging conditions.

Main Methods:

• Integration of Improved Sequential Variational Mode Decomposition (ISVMD) for signal reconstruction and feature extraction.
• Application of an Extreme Learning Machine (ELM) for trajectory prediction.
• Optimization of the ISVMD-ELM framework using the novel Red-billed Blue Magpie Optimizer (RBMO).

Main Results:

• The proposed RBMO-ISVMD-ELM approach demonstrated superior performance compared to state-of-the-art methods.
• The framework effectively reconstructs noisy target echo signals into robust feature sequences.
• High precision in predicting concealed ground target trajectories was achieved under demanding conditions.

Conclusions:

• The RBMO-ISVMD-ELM framework offers a robust and accurate solution for concealed target trajectory prediction in mmWave sensing.
• The synergistic combination of ISVMD, ELM, and RBMO enhances adaptability and computational efficiency.
• This approach significantly overcomes limitations posed by environmental interference and concealment.