Watermarking of Electronic Patient Record in Parkinson Disease Affected Speech: A Robust and Secure Audio Hiding Technique for Smart e-healthcare Application
Aniruddha Kanhe, Aghila Gnanasekaran · The interdisciplinary journal of Discontinuity Nonlinearity and Complexity · 2020
A cloud-based framework for security of electronic patient record, in telediagnosis of Parkinson's disease (PD) speech signal has been proposed in this paper. The security and authenticity of patient's personal information is achieved using proposed Discrete Cosine Transform-Singular Value Decomposition (DCT-SVD) based audio watermarking technique. The automatic classification of PD affected speech signal from healthy person's speech signal requires high computational accuracy and in this work, it is addressed by extracting various time frequency features for the Support Vector Machine (SVM) classifier. In the proposed framework the speech signal of PD suspected person is recorded by a mobile application. The personal information is securely embedded in this speech signal using proposed DCT-SVD based audio watermarking technique, before transmitting to cloud for automatic classification. To ensure the security of electronic patient record (EPR), proposed watermarking technique is tested against various signal processing attacks and the performance of classifier has been evaluated by computing classification sensitivity, specificity, accuracy and area under the curve of receiver operating characteristics (ROC). The theoretical proofs and experimental results, show that proposed framework provides telediagnosis of PD affect patients without compromising the security and privacy of patient's records. The classification accuracy of $89\%$ with high data payload of $6kbps$ is proposed leading to a smart and secure e-healthcare application.