Investigation of Biometric Date Input to Multi-Layer Perceptron
Kana Kurata, Chihiro Ikuta, Yoko Uwate, Yoshifumi Nishio · 2012
In this study, we propose a network model with input of bio-metric data to the hidden layer neurons of the Multi-Layer Perceptron (MLP). The injecting of the noise is known to be effective to the improvement of the performance of the MLP. We inject the biological pulse wave as a source of this noise. We investigate the performance of the MLP when the sam-pled data of biological pulse waves obtained from real bio-logical experiments are injected into the neurons. Computer simulated results show that the performance of the biological pulse wave is better than the case of random noise. Further, we investigated the characteristics of the proposed method, 1.