Step Counting via Pitch-Detected Signal Filtering

The patent segments the step counting process into multiple stages: obtaining triaxial acceleration signals, performing high-pass filtering to remove low-frequency components, conducting pitch detection to identify the fundamental frequency, applying low-pass or band-pass filtering to isolate the pitch component, and finally counting extreme value points. This segmentation transforms a simple but inaccurate peak-counting method into a multi-stage process that maintains simplicity while significantly improving accuracy.

The patent introduces intermediate processing steps between raw signal acquisition and step counting. Specifically, pitch detection serves as an intermediary that identifies the fundamental frequency of the oscillatory signal, and filtering operations act as intermediaries to remove noise and isolate relevant signal components. These intermediaries enable accurate step counting by transforming the raw acceleration signal into a form where extreme value points reliably correspond to actual steps.

The patent leverages the periodic nature of walkrun motion to simplify filtering operations. By performing pitch detection to identify the fundamental frequency of the periodic oscillatory signal, the system can apply low-pass or band-pass filtering with specific cutoff frequencies based on the detected pitch. This periodic action approach allows the use of relatively simple filtering operations tailored to the specific signal characteristics, rather than requiring complex universal filters.

The patent dynamically adjusts filtering parameters based on the detected pitch of the acceleration signal. The cutoff frequency of the low-pass or band-pass filter is set based on the pitch detection results, allowing the filter to adapt to different walkrun speeds and patterns. This parameter change strategy enables accurate filtering without requiring fixed complex filter designs, as the filter characteristics are optimized for each specific signal instance.