Biometrical and Psychophysiological Assessment through Biosensors

Pattern recognition is crucial in human interactions. It allows, for example, recognition of others’ identity and reactions to one’s actions. With the increasing use of machines, control systems and information technologies in everyday life, it has become advantageous for those systems to be capable of doing human biometrical and psychophysiological recognition. This work develops a framework methodology for the mentioned recognition through physiological data. For biometric recognition, a partially fiducial method using the electrocardiogram (ECG) is proposed. For the segmentation of the ECG signal into heartbeats, a fiducial method is applied in which the R peak is detected and used as alignment reference. For feature extraction, a non-fiducial method based on Principal Components Analysis (PCA) is applied. Finally, Euclidean distances and K-Nearest Neighbours (k-NN) classifier are used for user identification; the Euclidean distance is used for authentication purposes. Ultimately, 0% and 0.3% error rates are achieved for identification and authentication, respectively. In psychophysiological evaluation, two studies are presented – emotion evaluation and population group classification (drug abusers vs. control group) – using Blood Volume Pulse (BVP), electro-dermal activity (EDA) and respiratory signal (RESP) acquisitions. Fiducial feature extraction is performed and two classification methods are applied. The first uses PCA for conversion of the initial features to a feature space that maximizes their differences, followed by Euclidean distance and K-NN classification. The second used only Euclidean distance and K-NN classification. Emotion assessment from the available data was not successful; in contrast, the distinction between population groups was achieved with a 0% error rate.