An Analysis of Audio Features to Develop a Human Activity Recognition Model Using Genetic Algorithms, Random Forests, and Neural Networks

Table 5

Evaluation of each model.

HAR model	Accuracy	Number of samples	Number of activities
RF with 9 features	0.814	1,191	8
NN with 9 features	0.777	319	8
RF with all features	0.856	1,191	8
NN with all features	0.646	319	8
RF with all MFCC features	0.842	1,191	8
NN with all MFCC features	0.110	319	8
Kabir et al. scenario A	0.748	1,000	10
Kabir et al. scenario B	0.748	1,300	13
Kabir et al. scenario C	0.733	1,600	16