<![CDATA[Effective medical anamnesis for patients with hearing or speech impairments requires accessible communication channels. This study evaluates a vision-based hand gesture recognition pipeline to interpret gestures representing Common Cold complaints during healthcare triage. Hand landmark data comprising 63 spatial coordinates per frame were captured across three human subjects using a single RGB webcam powered by the MediaPipe Hand Landmarker under varying distances (1 m and 2 m) and ambient lighting conditions. The dataset covers seven gesture classes based on standard Indonesian Sign Language (SIBI): Saya (I), Sakit (Sick), Flu, Batuk (Cough), Demam (Fever), Pusing (Headache), and Nyeri Menelan (Painful Swallowing). Three supervised classifiers—K-Nearest Neighbors (KNN), Support Vector Machine (SVM), and Gaussian Naive Bayes (GNB)—were comparatively evaluated using a 5-fold Stratified Cross-Validation scheme. Experimental results demonstrate that Gaussian Naive Bayes achieved the highest classification performance with an overall accuracy of 0.81 and a macro F1-score of 0.81, followed by SVM (accuracy 0.80, F1-score 0.80) and KNN (accuracy 0.76, F1-score 0.76). These findings indicate that the continuous Gaussian distribution assumption aligns effectively with normalized knuckle coordinate features. Rather than serving as an automated disease diagnostic instrument, this framework functions as a reliable assistive interpreter for symptom-related sign language in clinical anamnesis.]]>
