This study focuses on the development of an efficient method by combining the feature extraction and classification methods and their implementation on the microcontroller test platform. The cosine Stockwell transform (CST) is used for extracting the significant amount of information from the corresponding ECG signals in lower dimensions. These features represent each of the ECG signals and are further identified using PSO-tuned twin support vector machines (TSVMs) into their different categories. The proposed method is implemented on the 32-bit advanced RISC machine (ARM) platform. The platform is validated on the benchmark MIT-BIH arrhythmia data generated in real time and evaluated under category-oriented analysis scheme. The platform is integrated with the Wi-Fi module which sends the information of classified outputs to a remote platform. Once an abnormality is detected by the platform, a pop-up message can be viewed on the displaying module interfaced with the platform which behaves as an alarm. The platform reported an accuracy of 95.8% in the category-oriented assessment scheme. Such type of prototyping of proposed method on hardware platforms deliver an assistive diagnostic solution to the users and should be employed in hospitals for cardiovascular disease diagnosis by providing an enriched platform capable of performing real-time diagnosis for telemedicine applications.

Chapter Contents:

• 5.1 Introduction
• 5.2 Methods
• 5.2.1 Stockwell transform (ST)
• 5.2.2 Twin support vector machines (TSVMs)
• 5.2.3 Particle swarm optimization (PSO)
• 5.3 Proposed method
• 5.3.1 MIT-BIH data
• 5.3.2 Preprocessing
• 5.3.3 R-wave localization and ECG segmentation
• 5.3.4 Feature extraction
• 5.3.5 CST feature recognition using TSVMs
• 5.4 Hardware implementation on Wi-Fi integrated embedded platform
• 5.4.1 Performance metrics
• 5.5 Results and discussion
• 5.5.1 Comparison with literature
• 5.6 Conclusion and future scope
• References

Inspec keywords: wavelet transforms; medical disorders; electrocardiography; telemedicine; medical signal processing; support vector machines; diseases; particle swarm optimisation; feature extraction; signal classification; wireless LAN; cardiovascular system

Other keywords: real-time ECG-processing platform; category-oriented analysis scheme; telemedicine applications; 32-bit advanced RISC machine platform; benchmark MIT-BIH arrhythmia data; category-oriented assessment scheme; PSO-tuned twin support vector machines; cosine Stockwell transform; cardiovascular disease diagnosis; hospitals; feature extraction; diagnostic solution; ECG signals; classification methods; TSVM; microcontroller test platform; real-time diagnosis; Wi-Fi module

Subjects: Optimisation techniques; Signal processing and detection; Computer communications; Integral transforms; Electrical activity in neurophysiological processes; Function theory, analysis; Local area networks; Digital signal processing; Biology and medical computing; Electrodiagnostics and other electrical measurement techniques; Integral transforms; Knowledge engineering techniques; Biomedical communication; Bioelectric signals; Optimisation techniques