Parkinson’s disease (PD) is a progressive neurodegenerative disorder characterized by motor symptoms including tremor, rigidity, and bradykinesia, with tremor-related movement disorders presenting a major clinical challenge for quantitative differentiation. Current diagnosis relies on subjective visual observation and clinical rating scales such as the Unified Parkinson’s Disease Rating Scale (UPDRS), Movement Disorder Society–UPDRS (MDS-UPDRS), and Hoehn and Yahr (H&Y) scale. Despite their clinical utility, these assessments remain time-consuming, prone to subjectivity, and limited sensitivity to subtle changes. To address these limitations, this study introduces a real time, contactless, computer vision-based system that can screen and differentiate tremors namely rest tremor (RT), essential tremor (ET), and cerebellar tremor (CT) through pose-based hand motion analysis. The proposed system adopts the combination of synchronized RGB and Time-of-Flight cameras to monitor the fine hand movements in real time during a standardized protocol involving the resting posture, postural arm extension, and finger to nose task. Three-dimensional hand movement is extracted using landmark tracking, specifically of the fingertips and processed to estimate the tremor frequency, Root Mean Square (RMS) displacement, and signal-to-noise ratio (SNR). These pose-specific kinematic features enabled statistically significant differentiation of RT, ET and CT, with significant group differences in RMS displacement, dominant frequency and SNR among the tremor groups is p < 0.001. Clinical validation was carried out on 400 samples collected from various participants recruited from AIIMS Mangalagiri following Institutional Ethics Committee approval (IEC No: AIIMS/MG/IEC/2025-26/389). The proposed system achieved high accuracy of 97.1% in differentiating the various types of tremors from one another in real time.