Abnormal Gait Detection and Classification Using Micro-Doppler Radar Signatures

Micro-Doppler radars have been used to accurately detect and classify human activities in various scenarios. This paper discusses the development of an X-Band micro-Doppler radar in detecting abnormal signatures embedded in micro-Doppler responses from activities such as walking, jogging, and jumping. To synthesize the condition of an abnormal gait, five test subjects were asked to perform the activities while wearing shoes, being barefoot, and wearing shoes accompanied by a heel lift insert in one shoe. Abnormal gait detection was performed using micro-Doppler responses of the various activities measured over two different look angles. The Short-Time Fourier Transform (STFT) was used to analyze the micro-Doppler characteristics along with Cadence Frequency Diagrams (CFD). Feature extraction was performed on the micro-Doppler responses for acquiring unique "hand-crafted" features commonly used in literature for gait analysis by micro-Doppler radar measurements. A secondary analysis utilized dimensionality reduction through Principal Component Analysis (PCA) and Linear Discriminant Analysis (LDA) to extract unique signatures and perform abnormal gait classification. An in-depth evaluation was performed on the different feature sets by the Weight K-Nearest-Neighbor (WKNN) and Support Vector Machine (SVM) classifier algorithms that determined the feasibility of discriminating between individuals performing the activities. The research allows for future determination of abnormal motion in specific activities via micro-Doppler response and machine learning that can further emphasize the ability of micro-Doppler radar to perform abnormal gait classification.