ABSTRACT
Abstract 176 7.1 Introduction 177 7.2 Unimodal and Multimodal Biometric Systems 178 7.3 Proposed Multimodal Biometric Authentication System 188 7.4 Methodology 192
7.4.1 Discrete Wavelet Transformation 192 7.4.2 Discrete Cosine Transformation 194 7.4.3 Singular Value Decomposition 196
7.5 Proposed System 197 7.5.1 Fingerprint Image Embedding in ECG Signal 197 7.5.2 Fingerprint Image Extraction from Watermarked
ECG Signal 198 7.5.3 Comparative Study between Original ECG Signal
Features and Watermarked ECG signal Features 199 7.5.4 Joint Feature Matching 199
7.6 Results and Discussions 199 7.6.1 Peak Signal to Noise Ratio 200 7.6.2 Correlation Coefficient 202
In recent times, identification authentication has been facing a number of challenges in verifying a person’s identity. Accurate verification is necessary, or else the person might suffer from identity crisis. Also, unauthorized access might result in misuse of personal information. To eliminate such discrepancies, many authentication processes have come into being. Of all the authentication techniques in use, the biometric authentication process that deals with human anatomical features provides the most reasonable solution. The work in this chapter proposes combining of electrocardiogram (ECG) and finger-print features to design a robust biomedical authentication system. According to the proposed methodology, after recording and extracting the ECG and fingerprint features, they are stored in templates in remote databases. On login attempt, the person’s fingerprint and ECG features are recorded to produce a watermarked ECG signal. The watermarked signal is transmitted over a communication channel and matched with the data stored in the database, at a remote feature matcher. Comparing the two, a match is found if the result obtained lies within a previously set threshold. To develop the watermarked ECG signal, the size of a gray fingerprint image, taken as the watermark, is calculated. The ECG signal recorded is converted to 2D signal as per the size of the fingerprint image. After decomposing the 2D signal by SWT into 4 subbands, DCT is applied to the subband Cd1. SVD is applied on the resultant image. SWT is also applied to the fingerprint image and decomposed into 4 sub bands, to each of which DCT is applied. Subsequently SVD is applied to the resultant image. Singular value of the 2D signal is modified using the singular value of Cd1 band of the watermark image. Finally IDCT and ISWT are applied to embed the watermark into 1D signal, which is
7.6.3 Structural Similarity Index Metric 202 7.6.4 Robustness Test of the Proposed Algorithm 209
7.6.4.1 Effect of White Gaussian Noise 209 7.7 Conclusion 210 References 211
subsequently reshaped into 1D signal. The method serves to be a strong authentication system.
