Abnormal Gait Behavior Detection for Elderly Based on Enhanced Wigner-Ville Analysis and Cloud Incremental SVM Learning

Table 2

Experiment data.

Category	Activity	Description	Falling down?	Binary accuracy (%)	Rank-1 recognition rates (%)	Rank-1 recognition rates (%)	Rank-1 recognition rates (%)
Normal gait behaviors	ADL	Normal walking	No	100	100	100	100
		Sitting on a chair then standing	No	100	94.7	97.2	98.6
		Squatting	No	100	95.2	97.1	98.3
		Climbing three steps	No	100	95.0	97.6	98.9
		Lying on a bed then standing	No	100	92.1	94.8	96.7
Abnormal gait behaviors	Forward fall	Ending on the knees	Yes	100	86.8	92.1	97.4
		Ending lying flat	Yes	100	88.3	95.3	98.8
		With recovery	No	100	86.3	93.8	97.5
	Backward fall	Ending sitting	Yes	95.8	87.4	93.5	96.5
		Ending lying	Yes	100	87.0	92.1	97.2
		With recovery	No	100	86.1	92.2	95.9
	Leftward fall	Ending lying	Yes	100	85.2	91.6	93.8
		With recovery	No	100	87.2	92.3	94.1
	Rightward fall	Ending lying	Yes	100	88.6	92.8	94.3
		With recovery	No	100	86.1	91.4	93.4
	Stumbling	Stumbling without fall	No	96.7	84.2	90.3	96.2
	Syncope	Slipping against a wall ending in sitting position	Yes	96.2	85.0	91.2	96.6