​ Abstract: Upper limb rehabilitation methods, such as the table sliding technique and pegboard exercises, are widely used because of their low physical burden on patients. A critical aspect of these methods is the ability to evaluate whether the exercises are appropriate for the patient’s condition and objectively assess the recovery progress. Traditionally, therapists have relied on direct observations and predefined evaluation criteria to assess movement performance and recovery. However, these approaches often require specialized equipment and technical expertise. To address these limitations, inertial measurement units (IMUs) have been proposed as a means of quantifying task performance and supporting clinical decision-making using objective data. Our previous studies, presented at SEIA’2024 and published in Sensors and Transducers, explored the application of IMUs in upper limb rehabilitation. However, challenges remain regarding sensor miniaturization and simultaneous multisensor measurements, owing to peg-size constraints. Recent advancements in time-of-flight (ToF) sensor technology have enabled the development of compact and cost-effective sensors. ToF sensors offer privacy advantages by not capturing direct images, making them suitable for healthcare applications. In this study, we compared two types of ToF-based sensors: one that detects object presence within an 8×8 pixel area based on height and another that uses 360° infrared light to measure object positions in 2D. We first applied support vector machine classification to the images obtained from both sensors, followed by classification using TensorFlow. The results were analyzed to evaluate the differences in the classification accuracy between the two sensor types and methods.

Keywords: Time of flight, LiDAR, Raspberry Pi, Upper limb rehabilitation, SVM, TensorFlow.

_________________________________________________________________________________________