Hip Motion Measurement and Classification Using Millimeter Wave Radar and Convolutional Neural Networks

After hip joint injury or replacement surgery, patients must undergo regular hip motion rehabilitation. This study seeks to replace the in-person doctor’s assessment with noncontact radar sensing. Multiple test subjects perform three hip motion patterns while being scanned by a millimeter wave radar. The micro-Doppler signatures are processed into a spectrogram image format for analysis. The different hip motions exhibit unique time-frequency features which are exploited by a convolutional neural network to classify the activities.