Accelerometry-Based Classification of Human Activities Using Markov Modeling

Table 3

Classification accuracy averaged over the twenty subjects available in the seven-activity dataset in the presence of spurious data. The values are reported as mean ± standard deviation.

Seven-activity dataset (spurious data present in proportion 1 : 3)
	M = 1	M = 2	M = 3	M = 4	M = 5
Without rejection of spurious data
GMM	63.3 ± 9.9	62.4 ± 11.4	60.2 ± 11.2	60.5 ± 13.0	56.0 ± 11.2
cHMM (1st level)	63.8 ± 9.3	61.4 ± 10.6	57.8 ± 10.9	55.0 ± 14.0	42.6 ± 13.9
cHMM (1st + 2nd level)	63.8 ± 9.3	61.8 ± 9.7	57.0 ± 12.5	50.1 ± 15.5	43.4 ± 14.6
With rejection of spurious data
GMM	85.4 ± 13.4	86.7 ± 14.9	83.4 ± 15.3	80.5 ± 16.2	81.7 ± 17.6
cHMM (1st level)	86.2 ± 12.9	87.1 ± 13.7	83.0 ± 14.6	80.2 ± 15.7	81.9 ± 17.6
cHMM (1st + 2nd level)	86.2 ± 12.8	86.1 ± 13.3	83.2 ± 14.0	76.1 ± 16.7	75.9 ± 20.6