An Intelligent Approach of Measurement and Uncertainty Estimation for Hidden Information Detection Using Brain Signals

• Published in: MĀPAN : journal of Metrology Society of India Vol. 37; no. 1; pp. 81 - 95
• Main Authors: Saini, Navjot, Bhardwaj, Saurabh, Agarwal, Ravinder
• Format: Journal Article
• Language: English
• Published: New Delhi Springer India 01.03.2022; Springer Nature B.V
• Subjects: Auditory stimuli; Brain; Crime; Electroencephalography; Feature extraction; Mathematical and Computational Physics; Mathematical Methods in Physics; Measurement Science and Instrumentation; Numerical and Computational Physics; Original Paper; Physics; Physics and Astronomy; Radial basis function; Simulation; Support vector machines; Theoretical; Uncertainty; Waveforms; Empirical mode decomposition; Support vector machine; Electroencephalogram; Feature ranking
• ISSN: 0970-3950; 0974-9853
• DOI: 10.1007/s12647-021-00493-7

The focus of the present work is to reliably determine the hidden information in the human brain accompanied by uncertainty estimation to develop an efficient computer-based hidden information detection procedure. A new approach is proposed to assess the concealed information in the individuals utilizing their single-channel electroencephalogram (EEG) waveforms based on experimental testing. EEG signals of 35 guilty and innocent participants were recorded corresponding to the pictorial and auditory stimuli with the help of a customized mock crime based protocol. Pre-processing of EEG data resulted in artifact-free average probe waveforms at Pz electrode location. Then, different time, frequency, wavelet, and empirical mode decomposition derived features corresponding to average pictorial and auditory probe waveforms were extracted. An optimal number of distinctive features were attained with the ReliefF ranking method. The proposed combination of seven auditory ranked features with a support vector machine (radial basis function) kernel was evaluated comprehensively to differentiate between the guilty and innocent subjects. The proposed combination achieved a sensitivity of 100%, the specificity of 87.50%, and the highest classification accuracy of 92.86%, using one EEG channel. The proposed arrangement performed better and exhibited low combined uncertainty than the competent feature extraction techniques in recognizing the concealed information.