Browsing by Author "Dr. M. Aksam Iftikhar"

Now showing 1 - 3 of 3

Fusion of Infrared and 3D Heatmap Volume Representation of Skeleton Data for Human Activity Recognition
(Library Information Services, COMSATS University Islamabad, Lahore Campus, 2025) Muhammad Taimoor Akmal; CIIT/FA22-RCS-025/LHR; Dr. M. Aksam Iftikhar; LHR TP 10031
Human Activity Recognition (HAR) is a fundamental component of intelligent monitoring systems in healthcare, rehabilitation, and assisted living, where accurate recognition of fine-grained medical actions is often challenged by subtle motion variations, illumination changes, and background clutter. This research investigates a multimodal HAR framework that integrates infrared (IR) video cues with skeleton- based motion representations to improve recognition reliability for medical-condition activities. A curated subset of the NTU RGB+D dataset is constructed using nine medical action classes (A41–A49: sneeze/cough, staggering, falling down, headache, chest pain, back pain, neck pain, nausea/vomiting, and fan self), comprising 1134 paired IR–skeleton samples with a defined training and validation protocol. The skeleton stream adopts PoseConv3D-style 3D heatmap-volume representations and is trained under two encodings (joint-based and limb-based). The IR stream is trained using an R(2+1)D-18 spatio-temporal backbone, and an ablation study demonstrates that full- frame IR training exhibits severe overfitting, whereas skeleton-guided subject-centric cropping improves generalization and reduces computational redundancy. Multimodal integration is performed through score-level fusion on a paired overlap validation subset, yielding improved recognition accuracy compared to single-modality baselines. Experimental results show strong skeleton-only performance and further gains through IR integration, with the best multimodal setting achieving 96.79% Top-1 and 100% Top-5 accuracy on the paired validation subset. The findings confirm that combining robust volumetric pose cues with appropriately processed infrared features improves discrimination among visually and kinematically confusable medical actions, providing an effective direction for multimodal HAR in healthcare-oriented environments.
Human activity recognition for musculoskeletal accidents at workplace
(Library Information Services, COMSATS University Islamabad, Lahore Campus, 2021) Hafiz Muhammad Abdullah; FA17-BSE-014; Dr. M. Aksam Iftikhar; LHR TP 7024
The industrial revolution that aims to automate industries without human interaction by using artificial intelligence. Employees are trained and are provided with standard operating procedures (SOPs) to work in a safe environment. However, sometimes due to the reckless behaviour of employees or improper training, accidents happen. Employers are held responsible for employee’s financial aid due to injuries or accidents that takes place at the workplace. This project will allow surveillance on employee’s activities for musculoskeletal injuries using Deep learning techniques for awkward posture recognition. The benefit of this project is that it will guide employers in training employees and will try to investigate the person responsible for the accident that takes place in the workplace.
Multi Layered Rule-Based Technique for Explicit Aspect Extraction from Online Reviews
(Library Information Services, COMSATS University Islamabad, Lahore Campus, 2020) Dr. M. Aksam Iftikhar; Mubashar Hussain; FA17-RCS-030; LHR TP 6402
Aspect is a feature of product, about which the user is expressing his/her opinion. In the field of sentiment analysis, the key task is to extract aspects or opinion targets from blogs, product reviews or discussions, users have expressed their opinions about a product. Extracting just positive or negative opinion is not much useful without knowing the targeted Aspect or Feature. If the aspect or feature is explicitly mentioned in the user‘s review, that aspect is called explicit aspect, otherwise it is termed as implicit aspect. Extracting explicit aspects from the opinion statements has been an active area of research. To achieve this milestone, multiple methodologies have been developed including supervised, rule-based and unsupervised approaches. Many researchers used dependency relation of the sentence also called syntactical dependencies to formulate the rules for aspect extraction. However this approach heavily depends on the authenticity of the employed POS tagger and dependency parser. Their outputs can be misleading as the users may not follow the grammatical rules strictly while writing their opinions. Another popular approach for explicit aspect extraction is sequential rules based approach, wherein the rules are formed by learning from user‘s behavior. However, in general, the sequential rule-based approaches have poor generalization capability. Moreover, existing approaches mostly considers the aspects as noun or noun phrase, so these approaches are unable to extract verb aspects. In this thesis, for overcoming the said problems with the existing approaches, we have proposed a multi-layered rule-based (ML-RB) technique using the syntactic dependency parser based rules along with some selective sequential rules. These rules are applied one after another (i.e. sequential rules after syntactic rules) for extracting noun explicit aspects form customer‘s reviews, thereby forming layers of the application. We also have crafted rules for extraction of verb aspects after rigorous analysis. In third layer, x we have used these verb rules, which are primarily based on the association among verb and opinion words. The proposed multi-layer technique compensates for the weaknesses of individual layers and yields improved results. As the proposed approach has capability to extract both noun and verb aspect, it has better Recall and F1-score on two publically available bench mark customer review datasets, as compared to other famous approaches