Monitoring joint mobility limitations (JML) for lower-limbs is essential for the early detection and management of musculoskeletal and neurological disorders that impact gait and functional. However, traditional joint range of motion (ROM) assessments are typically limited to clinical settings and lack the capability for continuous, real-world monitoring. This study explores the feasibility of monitoring joint mobility impairments using a wearable smart insole system. To validate the system, we designed a joint-specific motion protocol and collected walking data from four participants under normal and mechanically restricted conditions at the ankle, knee, and hip joints. All joint impairments were simulated on a single limb at a time using mechanical limiters, while normal walking trials served as baseline reference.  The collected data was segmented into two second windows and used to train a lightweight neural network model to classify the type of joint JML condition. Our results demonstrate the proposed system can effectively distinguish between normal and ROM impaired gait patterns. Under Leave-One-Subject-Out (LOSO) model evaluation setting, the proposed method achieved average classification accuracy of approximately 81% in distinguishing knee and ankle JML conditions against normal gait data from four participants. This research highlights the potential of low-cost, pressure-only wearable systems for continuous, personalized monitoring of joint function in everyday environments, laying the groundwork for future applications in remote rehabilitation and clinical decision support.