Human Activity Recognition (HAR) is becoming an essential part of human life care. Existing HAR methods are usually developed using a two-level approach, wherein a first-level Machine Learning (ML) classifier is employed to distinguish the static and dynamic activities, followed by a second-level classifier to identify the specific activity. These approaches are not suitable for wearable devices, due to the high computational and memory consumption. Our rigorous analysis of various HAR datasets opens up a new possibility that static or dynamic activities can be discriminated against through a simple statistical technique. Therefore, we propose to utilize a statistical feature extraction technique to replace the first-level ML classifier, thus achieving more lightweight computation. Next, we employ Random Forest (RF) and Convolutional Neural Networks (CNN) to classify the specific activities, achieving higher accuracy compared to the state-of-the-art results. We further reduce the computation and memory consumption of the above combined approach by applying pruning and quantizing techniques to CNN (PQ-CNN). Experimental results show the proposed lightweight HAR method achieved an F1 score of 0.9417 and 0.9438 for unbalanced and balanced datasets, respectively. On top of lightweight and accuracy, the proposed HAR method is practical for wearable devices by using a single accelerometer.

As societies age, the issue of falls has become increasingly critical for the health and safety of the elderly. Fall detection in the elderly has traditionally relied on supervised learning methods, which require data on falls, which is difficult to obtain in real situations. Additionally, the complexity of integrating deep learning models into wearable devices for real-time fall detection has been challenging due to limited computational resources. In this paper, we propose a novel fall detection method using unsupervised learning based on a denoising long short term memory (LSTM)-based convolutional variational autoencoder (CVAE) model to solve the problem of lack of fall data. By utilizing the proposed data debugging and hierarchical data balancing techniques, the proposed method achieves an F1 score of 1.0 while reducing the parameter count by 25.6 times compared to the state-of-the-art unsupervised deep learning method. The resulting model occupies only 157.65 KB of memory, making it highly suitable for integration into wearable devices.

Mild cognitive impairment (MCI) is an early stage of Alzheimer’s disease (AD), which is currently incurable. Early diagnosis of AD is essential for effective intervention since the World Alzheimer’s Report 2015 predicted the number of cases will triple by 2050. The 18F-FDG PET imaging technique, although effective in detecting metabolic activities in the brain, faces challenges such as low signal-to-noise ratios and limited data availability, which complicates the existing methods to extract necessary lesion information effectively for diagnosing early-stage MCI. To overcome these challenges, we introduce a novel deep learning-based model, ResGLPyramid, that combines convolution operations, MobileViTv3, and a global-local attention module (GLAM) block, to capture local and global representations. By utilizing a softened cross-entropy (SCE) objective function, the model reduces overfitting, improves generalization, and enhances the detection of subtle metabolic changes. The proposed model enhances the sensitivity and specificity of Alzheimer’s detection by leveraging local- and long-range interactions among critical diagnostic features that lead to more precise and efficient analyses. Experimental results show the ResGLPyramid model achieved an accuracy of 92.75% in classifying MCI and AD individuals, which is 3.44% higher than state-of-the-art methods.

Real-time detection of Covid-19 has definitely been the most widely-used world-wide classification problem since the start of the pandemic from 2020 until now. In the meantime, airspace opacities spreads related to lung have been of the most challenging problems in this area. A common approach to do on that score has been using chest X-ray images to better diagnose positive Covid-19 cases. Similar to most other classification problems, machine learning-based approaches have been the first/most-used candidates in this application. Many schemes based on machine/deep learning have been proposed in recent years though increasing the performance and accuracy of the system has still remained an open issue. In this paper, we develop a novel deep learning architecture to better classify the Covid-19 X-ray images. To do so, we first propose a novel multi-habitat migration artificial bee colony (MHMABC) algorithm to improve the exploitation/exploration of artificial bee colony (ABC) algorithm. After that, we optimally train the fully connected by using the proposed MHMABC algorithm to obtain better accuracy and convergence rate while reducing the execution cost. Our experiment results on Covid-19 X-ray image dataset show that the proposed deep architecture has a great performance in different important optimization parameters. Furthermore, it will be shown that the MHMABC algorithm outperforms the state-of-the-art algorithms by evaluating its performance using some well-known benchmark datasets.

Human Activity Recognition (HAR) has garnered attention as a significant technology that can enhance the quality of human life. However, existing HAR works still face great challenges such as a shortage of labeled data and the difficulty of rebuilding a deep learning model whenever the application environment (e.g., user or sensor position) changes. To address these challenges, we propose a new data-centric approach for HAR by using a Semi-supervised Generative Adversarial Network (SGAN). To improve the accuracy of HAR, we propose a data supplement strategy that systematically improves data quality, rather than the model, by using data refinement and data-driven feature extraction techniques. The proposed HAR method applies simple SGAN to achieve considerably high accuracy with only a small fraction of the labeled data. Therefore, the proposed HAR method can reduce overhead from data labeling, which is a labor-intensive and time-consuming process for many HAR tasks. Moreover, the data-centric HAR method is robust even in scenarios when there is a change in person/sensor location. Experimental results show that our method improves accuracy by as much as 3% over state-of-the-art semi-supervised HAR methods with only 3% of the data being labeled, leading to comparable accuracy to state-of-the-art HAR methods based on supervised learning.