Table 2 Classification performance for the LOSO experiment on the SData set. For each predicted label (tremor, FMI and PD) we report the average sensitivity, specificity and precision across 10 independent experimental trials. The first two rows present the predictive performance of standalone symptom classifiers. Such models arise during the initial pre-training step, when the separate branches that comprise the total model are separately trained against a single symptom label (the pre-training procedure is described in detail in Section Training details). The rest of the table, presents the performance of the fused multi-label classifier.

From: Unobtrusive detection of Parkinson’s disease from multi-modal and in-the-wild sensor data using deep learning techniques