| Summary: | In the current scenario, it is significant to design active learning paradigms for analyzing human activities using Wearable Internet of Things (W-IoT) sensors for health parameter analysis. Further, in the healthcare sector, data collection using decision-making tools uses wearable sensors for monitoring using Cloud assisted Internet of Things (IoT). Although several conventional algorithms and deep learning models show promising results in sensor data analysis for recognizing human behaviors, the evaluation of their ambiguity in decision-making is still difficult and several conventional systems are more complex. Due to the restricted computing capacity, low-power W-IoT devices need an optimized network to manage the healthcare data effectively and efficiently for reliable analysis. Hence, a new Human Activity Recognition based on Improved Bayesian Convolution Network (IBCN)has been proposed which allows each smart system to download data via either traditional Radio Frequency (RF) communication or low power back dispersion communications with cloud assistance. In IBCN, A distribution of the model's latent variable is designed and the features are extracted using convolution layers, the performance of the W-IoT has been improved by combining a variable autoencoder with a standard deep net classifier. Furthermore, the Bayesian network helps to address the security issues using Enhanced deep learning (EDL) design with an effective offloading strategy. The experimental results show that the data collected from the wearable IoT sensor is sensitive to various sources of uncertainty, i.e. aleatoric and epistemic, as especially named noise and reliability. Furthermore, lab-scale experimental analysis on patient's health data classification accuracy has been considerably developed using IBCN than conventional design as namedCognitive radio (CR) learning, deep learning-based sensor activity recognition (DL-SAR) and Cloud-assisted Agent-based Smart home Environment (CASE).
|
|---|
