Design of a Python-based Lecture Capture System using Haar Cascade Algorithm with Noise Reduction as Facility for Learner-Centeredness in OBE

Abstract

Lecture recording is a common system that higher-education institutions use as a tool to deliver quality education. This study focuses on designing an automated lecture capture system that would record the instructor's lecture using a face-detecting camera that will pan and zoom as the instructor moves and enhance the audio quality through noise reduction. Student surveys are implemented to quantify the effectiveness of a student's engagement on using the lecture capture system. Testing is done in a classroom of 36 students in a basic electronics course class in Mapúa University. The Haar Cascade algorithm is an algorithm that has a high positive detection rate with a low false positive detection rate and can process image quickly. Results show that the designed Lecture Capture System can detect, follow, and zoom in on the speaker. The noise reduction technique using the Soft Mask filtering improves the audio recorded by the system. The best audio comes from the lapel with AGC with an SNR of 42.9dB. The responses from the self-assessment surveys handed out to the students in the class had an overall mean value of 3.9 showing that the Lecture Capture System is a convenient educational resource and has a positive impact on their learning and study habits.