The proposed EBiLSTM-SOA model for heart rate classification achieved 24.70% higher accuracy and 19.44% higher specificity compared to traditional approaches.
Does the CAM-HRC model with EBiLSTM-SOA improve heart rate classification accuracy from PPG signals compared to traditional methods?
The proposed EBiLSTM-SOA deep learning model significantly improves the accuracy and specificity of heart rate classification from wearable PPG signals compared to traditional methods.
Effect estimate: 24.70% higher accuracy, 19.44% higher specificity
Cardiac activity monitoring is important for assessing cardiovascular status and supporting computational analysis of heart-rate pattern variations from wearable PPG signals. The development of wearable devices and public datasets has made continuous and noninvasive monitoring more viable. However, there are still problems that may impair the effectiveness of the technique such as motion artifact, signal noises, changes in the physiological states, and differences in sensor quality. Moreover, it is hard to create algorithms that would be highly accurate in the real-life scenario and would be able to generalize among diverse populations. These problems should be resolved in order to produce reliable, robust, and computationally consistent heart-rate pattern classification frameworks. Hence, this research paper proposes a cardiac activity monitoring-based Heart Rate Classification (CAM-HRC) model with the help of an intelligent deep learning approach. The data was first obtained on the standard benchmark sources referred to as WildPPG: A Real-World photoplethysmography (PPG) Dataset. Preprocessing at the baseline level in the collected data was obtained through data cleaning and normalization techniques. The Temporal Attentive U-Net (TAU) method was used to divide these pre-processed data into segments. These features were then extracted from segmented data with Self-Supervised Multi-Encoder Autoencoder (MEAE) method. The last stage was the classification of these extracted features with CAM-HRC model and the newly added Enhanced Bidirectional Long Short-Term Memory (EBiLSTM). The parameter tweaking of the traditional BiLSTM model was done with the help of a nature-inspired optimization algorithm known as Salamander Optimization Algorithm (SOA). The objective function of the whole process of proposed CAM-HRC model is taken as accuracy maximization. The suggested EBiLSTM-SOA model will divide the final heart rate output into four classes, normal heart rate, Tachycardia, Bradycardia and Rhythm Irregularity. The accuracy and specificity of the proposed EBiLSTM-SOA model of the CAM-HRC are 24.70 per cent and 19.44 per cent higher than the other methods of traditional approaches, respectively.
B. Sabitha (Mon,) conducted a other in Heart rate classification. CAM-HRC model with Enhanced BiLSTM (EBiLSTM) and Salamander Optimization Algorithm (SOA) vs. Traditional approaches was evaluated on Accuracy and specificity of heart rate classification (24.70% higher accuracy, 19.44% higher specificity). The proposed EBiLSTM-SOA model for heart rate classification achieved 24.70% higher accuracy and 19.44% higher specificity compared to traditional approaches.