Patient-Aware EEG-Based Feature and Classifier Selection for e-Health Epileptic Seizure Prediction

On the wavelet-based compressibility of continuous-time sampled ECG signal for e-health applications

“…PRD performances regarding the number of wavelets, K; retained by the OMP for the 75 test signals and for L; D ð Þ equal to (2, 64), (2, 128), (3,64) and (3,128) are measured and plotted. However, they are not included in this paper for sake of clarity.…”

“…In fact, x could be compressible regarding Wleading to a coefficients' vector s in which the total number of significant coefficients, K, is small compared to the length, D; of s. In this case, s can be transformed in a sparse vector by keeping the largest K coefficients and setting others to zero. Besides, the coefficients a i , especially the K largest ones, present relevant signal features that are involved in machine learning algorithms [3,6].…”

“…In such situation, a wearable and connected ECG healthcare device is required as it increases user's comfort by allowing his mobility. Wireless monitoring is also adopted in remote diagnosis systems, which are proposed as a solution to healthcare issue for lowincome, hard-to-reach or rural zones [2,3]. It is recently involved in health crisis such as Coronavirus Disease 2016 (Covid-19) pandemic either to monitor patients on lockdown or to free up beds for the sickest Covid-19 patients [4].…”

“…In addition to patient monitoring, the ECG is recently involved in innovative biomedical applications such as decision support system to help in the diagnosis of ambiguous cases or as a biometric signal for authentication and identification [5,6]. The latter solutions involve the machine learning algorithms for classification and utilize intrinsically the wavelet transform as a feature extractor [2,3,[5][6][7][8].…”

“…Besides, the system would benefit from higher compression ratios, lower complexity and lower power consumption than those reached by using only the LC-ADC [26]. To the best of the authors' knowledge, while many works have been interested in compression of discrete-time sampled data [12,19,20], no study to date is reported about compression of continuous-time sampled ECG with the DWT and by applying threshold [3,6,17,19,20]. Thus, in this paper, the authors propose to extend their DWT-based compression study of continuous-time sampled electrocardiogram data to achieve better CRs.…”