Table 1.

MCPD-Net vs. unimodal architectures. Our proposed multimodal method outperforms all of the other approaches.

		Precision	Recall	${\mathit{F}}_1$-Score
Silhouette (Sil)	CNN	0.17	0.40	0.24
	VAE	0.49	0.49	0.47
	RF	0.46	0.39	0.41
	LSTM	0.45	0.40	0.41
Accelerometer (Acl)	CNN	0.53	0.45	0.44
	VAE	0.63	0.55	0.44
	RF	0.59	0.45	0.43
	LSTM	0.58	0.47	0.42
MCPD-Net		0.71	0.77	0.66