Article ID: JJPTF_2023_15
<Introduction> A Freeze Index (FI), calculated from acceleration, is reported objective assessment of freezing of gait. In this study, we investigated the attachment site of the accelerometer and the window width of FI from a feature selection perspective to explore effective measurement methods for FI. <Methods> We included 7 patients with Parkinson’s disease. Accelerometers were attached to the third lumbar spinous process and the distal part of both shins. FI was calculated by STFT analysis. FI was calculated using a single sensor (lumbar, right shin, left shin) and multiple sensors (both shins, lumbar and both shins) at 6 different window widths (1-6 second). We ranked each FI value based on mutual information and Euclidean distance by feature selection. <Results> The top selected features based on both indicators were the FI calculated using a single sensor (right shin, left shin, and lumbar) with a window width of 6 second and the FI calculated using the lumbar and both shins (max) with a window width of 1 second. <Conclusion> This study suggested that a 6-second window width FI by a single sensor is beneficial.