A Nonfiducial PPG-Based Subject Authentication Approach Using the Statistical Features of DWT-Based Filtered Signals

Table 3

Performance comparison of the proposed approaches with methods in the literature.

Method	Subjects	Accuracy (%)
Fiducial features [20]	17	94
Fiducial features of 1^st & 2^nd derivatives [24]	30	94.44
Dynamical system model [25]	23	95
Clustering+Boltzman Machines+Deep Belief Networks [27]	12	96.1
Convolutional neural network [30]	43/20	78.2/83.2
Four-layer-deep neural network [33]	22	96
Proposed approaches	42
Augmented features from time domain signal		97.89
Augmented features from the four bands		99.3