Application of de-shape synchrosqueezing to estimate gait cadence from a single-sensor accelerometer placed in different body locations
Objective: Commercial and research-grade wearable devices have become increasingly popular over the past decade. Information extracted from devices using accelerometers is frequently summarized as "number of steps" or "activity counts". Raw accelerometry data that can be easily extracted from accelerometers used in research, for instance ActiGraph GT3X+, are frequently discarded. Approach: In our work, we use the raw data recorded from a single sensor installed in different body locations to extract gait cadence and other gait characteristics via an innovative use of the de-shape synchrosqueezing algorithm. The proposed methodology is tested on data collected in a semi-controlled experiment with 32 participants walking on a one-kilometer predefined course. Walking was executed on a flat surface as well as on the stairs (up and down). Results: With the leave-one-subject-out cross validation, the accuracy and F1 of determining if a subject is walking from the single sensor installed on a wrist (hip, left ankle, right ankle respectively) was found to be 86 respectively). The cadences of walking on a flat surface, ascending stairs, and descending stairs, determined from the wrist (hip, left ankle, and right ankle, respectively) sensor, were 1.98+-0.15 Hz, 1.99+-0.26 Hz, and 2.03+-0.26 Hz respectively (1.98+-0.14 Hz, 1.97+-0.25 Hz, and 2.02+-0.23 Hz, respectively; 1.98+-0.14 Hz, 1.93+-0.22 Hz and 2.06+-0.24 Hz, respectively, and 1.98+-0.14 Hz, 1.97+-0.22 Hz, and 2.04+-0.24 Hz, respectively), which indicates that the cadence is fastest while descending stairs and slowest when ascending stairs. The larger standard deviation observed on the wrist sensor is in line with our expectations. Conclusion: We show that our method can detect walking bouts and extract the cadence with high accuracy, even when the sensor is placed on the wrist.
READ FULL TEXT