In sports dance (SPD) education, it is common to give excessive value on technique while ignoring emotion. This issue causes emotions and motion to be poorly integrated, which negatively impacts the educational result. As a result, this study employs a sensor to gather video data from SPD performers and then extracts the crucial characteristic spots to estimate their poses. This study proposes unique capuchin search-driven bi-directional tuned recurrent network (CS-BRN) architecture to efficiently categorize dancers' emotions. Additionally, the arousal value can be used to categorize the arousal valence (AV) emotional paradigm into various parts. The AV paradigm performs superior in high and low fields, which correlate to exciting, nervous and mild feelings, respectively. We employed metrics to evaluate this study's performance like precision (90%,) accuracy (96%), F1-score (92%), recall (91%), The findings of the trial demonstrate that the suggested method has a high degree of emotional recognition accuracy and can precisely identify the crucial moments in the performers' technical motions. This study significantly aided in the emotional identification and alleviation of SPD participants during the education procedure by precisely identifying the crucial parts of their technical motion presentations.