Fittology: a web-based AI pose correction workout tool for beginners with audio-visual feedback using raspberry PI4. 6

Undergraduate Thesis: (Bachelor of Science in Computer Engineering) - Pamantasan ng Lungsod ng Maynila, 2024. 56

5

ABSTRACT: STATEMENT OF THE PROBLEM: Newbie fitness enthusiasts often struggle with limited access to professional guidance, difficulties in maintaining correct workout postures, and the absence of exercises for beginners. These challenges can be discouraging and hinder progress. In response to these issues, this study focuses on the development and evaluation of Fittology, a Web-Based AI Pose Correction Worout Tool for Beginners with Audio-Visual Feedback Using Rasberry Pi4, which seeks to revolutionize the way people approach fitness and exercise RESEARCH METHODOLOGY: The methodology of data gathering used for this study is the quantitative method, which uses descriptive design. The first step of Fittology is the data collection process wherein the user's data is collected (fitness goal, existing medical conditions, and any injuries). This information is vital for proceeding to the AI Pose Correction Tool. SUMMARY OF FINDINGS: The main objective of the study is to develop a web-bsed AI workout tool that uses pose estimation technology that provides feedback after each exercise for posture correction during exercise routines, aimed at helping beginners achieve fitness goals. To gather data, the researchers utilized a pre-survey and post-survey for workout beginners and separate assessment for the fitness instructors through Google Forms. The researchers used convenience sampling and various statistical analysis from the chosen participants to satisfy the objectives and scope of the study utilized the ISO/IEC 25010 quality model and Likert Scale questions to measure data from the respondents. The researchers conducted a pre-survey for the workout beginners and gym-goers. The pre-survey found significant obstacles to attending a traditional gym. The most common reason given (97%) was a lack of confidence, indicating that many people felt self-conscious about their appearance or were concerned about being judged by others. Additional reasons included a lack of time (95%) and a lack of budget (93%), suggesting that time constraints and financial barriers also play a role in preventing people from regularly attending the gym. These findings highlight the need for alternative fitness, such as the web-based workout tool we propose, which can offer a more accessible approach for those facing these common challenges. The researchersalso evaluated the usability and effectiveness of our web-based AI pose estimation tool, Fittology, with input from fitness instructors. This was done with Google Forms. The results showed that Fittology provides accurate pose detection and consistent feedback during workouts. Instructors rated the system's ability to capture exercise poses and identify correct body alignment as effective to very effective (most answered 1-2), indicating that the tool is capable of guiding users toward better exercise form. Furthermore, Fittology's role in improving exercise techniques was viewed positively, with a unanimous I result for the system's ability to address needs of the user. Reading user satisfaction, the fitness instructors found Fittology's interface to be user-friendly and easy to navigate, with responses indicating high levels of satisfaction with its overall usability and functionality. The system's simplicity and intuitive design earned praise, and most instructors stated they would recommend Fittology to their peers and clients. These findings suggest that Fittology holds as a solution for those looking to improve their fitness without the need to attend a traditional gym, offering both effective guidance and a user-friendly experience. The researchers also conducted a post-survey. Based on the responses from 100 users who utilized Fittology, a web-based AI pose estimation workout tool, the results provide valuable insights into user satisfaction, system efficiency, and overall functionality. In terms of future use, most respondents (48%) indicated that they may consider using it but they are not entirely sure. With regards to pusture correction, 65% responses of very helpful took the lead. A leading 45% found the visual feedback feature very helpful. For the audio feedback feature, only 37% found it helpful. Lastly, a leading 73% of respondents would recommend Fittology to others, Essentially, in regards with the overall user evaluation in the post-survey, majority of the users fount it user-friendly (41%) and had an overall pleasant time utilizing Fittology (57%). Along with the post-survey was a system functionality form. This focused more on the technical aspects of the system and uses metrics of the ISO/TEC 25010 software quality model. The results show a positive response regarding the features of the system, which are website functionality (78%), pose estimation accuracy (74%). This also includes functional suitability (76.3% average of 4 responses), and performance efficiency (73% average of 4 responses.) CONCLUSION: In conclusion, Fittology, a web-based AI pose correction workout tool designed for beginners with audio-visual feedback using raspberry pi 4, demonstrates effective tracking and corrects user posture during workouts. The pose estimation technology provides precise feedback on exercise form, which contributes to more effective workouts for beginners. This is also supported by the fitness instructors and user evaluations, showing that Fittology accurately captures exercise poses and guides users toward better posture and alignment. The acceptance level among fitness professionals and trainers for Fittology is generally positive. Survey results from fitness instructors indicated they found Fittology's pose estimation feature to be good for the body alignment in exercising. Additionally, they rated the system's usability and functionality highly, with many indicating they would recommend Fittology to their colleagues and clients. This positive response suggests that Fittology meets the expectations of fitness experts, reinforcing its effectiveness as a workout tool. The system's ability to capture exercise poses accurately and provide real-time feedback contributes to its approaval among fitness. Moreover, Fittology has proven effective in assisting workout beginners, as seen on the post-survey results. The system is proven to be user-friendly, and an overall good experience based on the respondent's answers. Users also reported high levels of satisfaction with the system's pose estimation and audio-visual feedback features. Although, there are some mixed opinions which could point to possible areas of improvement for the system. Most respondents also noted that Fittology played a role in helping them maintain proper form and technique during workouts. This feedback suggests that Fittology provides valuable support to workout beginners as they reach their fitness goals with confidence in the comfort of their homes. Overall, Fittology offers an effective and accessible alternative to traditional gym-based workouts. Its AI pose estimation technology and audio-visual feedback features keep the system engaging for the users. This is coupled with its user-friendly interface and approval among fitness professionals. It proves to be a promising tool for gym beginners. Through there are some imperfections such as environmental camera distractions and instability, the results of this study support the capability of Fittology as a helpful workout tool, as it allows users to achieve their beginner goals in a home-based setting.

5