Human Activity Recognition Based on a Modified Capsule Network

Table 8

A comparison of the proposed approach with other studies on WISDM.

Datasets	Model	Accuracy (%)
WISDM	DNN [23]	90
	Capsule and LoRa [29]	95.2
	1D-HARCapsNet [30]	98.67
	CapsGaNet [31]	96.8
	2D CNN [43]	89.67
	Multichannel CNN-GRU [44]	96.41
	LSTM-CNN [45]	95.85
	Multi-input CNN-GRU [46]	97.21
	DCNN [47]	94.18
	Multihead convolutional attention [48]	96.4
	MCN (this paper)	98.21