Gait Analysis Using Waist-mounted Accelerometer Based on SVM Classifier

• Main Authors: YU, TUNG-HUA, 游東樺
• Other Authors: WU, CHAO-CHENG
• Format: Others
• Language: en_US
• Published: 2019
• Online Access: http://ndltd.ncl.edu.tw/handle/yd6859

碩士 === 國立臺北科技大學 === 電機工程系 === 108 === Gait is an important pattern of movements for animal with limbs. In neurology science, gait always plays an important role. It provides much information jointly from nerves and muscles. The gait disorders have three main aspects, which are sensory deficits, myelopathy and parkinsonism. This thesis focuses on the behavior of patients with myelopathy. This thesis took advantage of tri-axial accelerometers as measurement equipment of gait because it is easily available in consumer market with cheap price. Besides, tri-axial accelerometers also provides high sampling rate to trace tiny movement. The sensor was mounted on volunteer waist to trace change of center of body gravity. The signals of gaits from tri-axial accelerometers were processed and transformed from time to time-frequency domain through Hilbert-Huang Transform (HHT). The observation window of accelerometer data was separated into several gait cycles, which are stance phase, swing phase, double support, step and stride. Statistical behavior for each gait cycle was calculated as the features for classification. Then, Orthogonal Subspace Projection, Fisher score, and mutual information were utilized to prioritizes important features. It is normally difficult to collecting a large number of gait data from Cervical Spondylotic patients. Due to this constraint, this thesis utilized support vector machine (SVM) as the classifier. The experimental studies demonstrated that the proposed method offers a good classification performance with small amount data.