Abstract
Sleep apnea syndrome (SAS) is a common sleep disorder and chronic disease in which people's airway becomes partially or completely blocked during sleep and each pause in breathing is called a sleep apnea event, can last from ten seconds to minutes. In general, people are difficultly conscious of SAS and have to diagnose with an overnight sleep test, which is called a polysomnography (PSG) in the sleeping center of the hospital. However, PSG examination is a labor-intensive, limited environment, uncomfortable, and high-cost examination. Therefore, this study implements tri-axial accelerometer (TAA), which is a low-cost and convenient monitoring system. The two group of TAA attached to the left side of thorax and abdomen and can be used to capture the thoracic (THO) and abdominal (ABD) movement signals. The principal component analysis (PCA) algorithm used for transfering the three-dimensional raw data sensing from TAA to one-dimensional respiratory signals. The proposed modified PCA algorithm (modified PCA) can improve on the effect of the transformation, which is better than PCA algorithm, and signals are more stable when comparing with PSG. This thesis proposed modified PCA to obtain the respiratory waveform from TAA compared with PSG can reach 90\% similarity and detection result can reach closely 80\% Identification. The event of CSA and HYP still can not distinguished well and the other physiological signal or feature must be incorporated to enhance the identification.