Parkinson's Disease (PD), the second most common neurodegenerative disorder, is projected to increase by 60% by 2037. Treatment titration (e.g., medication dosage, deep brain stimulation (DBS) parameterisation) requires precise tracking of motor symptoms (e.g., tremor, bradykinesia) to reduce disability and improve quality of life. Out-of-clinic monitoring is critically needed for optimal outcomes. We introduce PDTrack, a comprehensive telemedicine system for routine, home PD assessment. PDTrack consists of: 1) a novel wearable combining accelerometer, gyroscope, and muscle activation (mechanomyography (MMG)) sensors; 2) a tablet running customised apps that guide patients through home PD assessment; and 3) heterogeneous motion/myography fusion algorithms for PD symptom classification. We hypothesise that fusing movement and myography features, rather than purely inertial features, improves symptom classification. PDTrack has been deployed in the homes of 23 people with Parkinson's (PwP; age 63.02±7.49; female: 8) who regularly use the system without clinical supervision. PDTrack achieved 72.2% accuracy in diagnosing bradykinesia and tremor compared with a clinician's video assessment using the Unified Parkinson's Disease Rating Scale (UPDRS). MMG enhanced accuracy (+5%) for some symptoms, notably postural tremor. Joint Mutual Information (JMI) analysis highlights dominance of frequency-based features across all classifications, followed by signal regularity features, which are more prominent in dynamic (e.g., kinetic tremor, bradykinesia) than static tasks (e.g., rest/postural tremor). Coherence was significant only for postural tremor. PDTrack is easy to operate, requires no clinical training, and can be used independently by patients for regular, 'on-demand', PD assessment. Results provide a basis for broader homebased deployment to monitor progression and titrate treatment.