An Adaptive Pooling YOLO-based Multimodal Detection Algorithm for Secure Automated Interview Assessment

The current recruitme?nt system requires efficient remote technical intervi?ew processes that provide secure and scalable solutions while? reducing bias. Traditional int?erview methods a?re? time-consu?ming and often produce inconsistent results due to? their reliance on human ju?dgment. Online recruitment has also intr?oduced challenges such as candidate authentication, cheating detection, and effective evaluation of technical skills. Advancements in Artificial? Intelligence? (AI) and Deep Learning (DL) have enabl?ed automated assessment systems with improved decision-making, contextual unders?tanding, and real-time interaction capabilities. This research proposes an automated interview framework t?hat integ?r?ates Natural Language Processing (NLP), computer vision, and speech technologies t?o conduct secure technical evaluatio?ns based on job requirements. The system dynamically generates interview questions using GPT-based language understanding and monitors candidates through an improved AP-YOLO model for real-time mon?itoring an?d? cheating de?tection. A CNN model trained on the FER2013 dataset is used for f?acial emotion recognition? to support effectiv?e evaluation while maintain?ing candidate confidence. The framework enables natural interaction t?hr?ough spee?ch-t?o-text and text-to-speech technologies while ensuring security th?rough browse?r activity monitoring, multiple-person detection, and vi?sual object recognition. The proposed system improves recruitment efficiency by automating scoring, generating adaptive questions, a?nd providing unbiased evaluation withou?t human intervention, while future work will focus on semantic answer eval?uat?ion, multilingual support, and advanced beha?vioral analysis to enhance AI-based de?cision ma?king?.