Continuous Gesture Sequences Recognition Based on Few-Shot Learning

Table 4

Influence of linear and nonlinear dimensionality reduction methods on model recognition accuracy.

Dataset	Similarity measure	PCA; 4-way 1-shot	PCA; 4-way 3-shot	Autoencoder; 4-way 1-shot	Autoencoder; 4-way 3-shot
RWTH	MSE	73.48%	74.82%	79.45%	80.72%
	Cosine	68.67%	70.16%	89.73%	90.16%
SMD	MSE	74.62%	74.47%	91.14%	91.11%
	Cosine	71.53%	74.13%	92.17%	93.23%