Gait analysis device and computer program product

A gait analysis device for observing a walking condition of a subject and preventing the subject from falling, the device comprising:

• a hardware processor;
• an acceleration sensor;
• and
• a memory for storing processor-executable instructions that, when executed by the processor, cause the processor to control the gait analysis device to at least:
• receive acceleration data from the acceleration sensor indicative of acceleration of the subject at or near a waist of the subject;
• calculate a variance of the acceleration data in each first setting time period;
• determine sequential variances are continuously equal to or less than a threshold value for a second setting timing period or longer, and thereafter the sequential variances continuously exceed the threshold value for a third setting time period or longer, and determine, as a walking start point, a time at which acceleration data corresponding to a last one of the variances equal to or less than the threshold value is obtained, the second setting time period and the third setting time period being both longer than the first setting time period;
• and
• estimate a walking condition and a fall risk of the subject subsequent to the walking start point by at least:
• calculating a feature quantity of the received acceleration data of the subject for a transitional time period that starts from the walking start point and calculating a variation based on the feature quantity, the calculated feature quantity including at least an average value of the acceleration data in a horizontal direction orthogonal to a moving direction of the subject for the transitional time period;
• and
• estimating the risk that the subject will fall based on the calculated feature quantity and the calculated variation, the estimating using a pattern recognition algorithm that is based on a learned relationship between class labels indicative of different risk levels and values of the feature quantity and variation based thereon for test subjects while walking.