A JS Cadence Sensor

signal-processing

Published

June 2, 2022

At some point during the pandemic, I became suspicious about the accuracy of my Saris H3 cadence sensor, so I wrote my own. (It turns out that the cadence sensor is fine.)

RPM:

How it works

The algorithm is embarrassingly simple for how well it works:

Capture the time series of the accelerometer data (with a 1s half-life exponential decay),
compute the PCA, and extract the first principal component. You now have a 1D time series
Compute the FFT, and find the period with largest energy.

Smartphones: supercomputer processing power, tricorder sensors. Amazing stuff.

---
title: A JS Cadence Sensor
categories: [signal-processing]
date: 2022-06-02
---

At some point during the pandemic, I became suspicious about the accuracy of my [Saris H3](https://saris.com/products/h3-direct-drive-smart-trainer) cadence sensor, so I wrote my own. (It turns out that the cadence sensor is fine.)

<script src="../src/lib/d3.v5.js"></script>
<script src="../src/lib/dsp.js"></script>

<button id="start">Start Cadence Sensor</button>
<div>RPM: <span id="rpm-reader"></span></div>
<div id="error-message"></div>

<div id="chart1"></div>
<div id="chart2"></div>
<script src="cadence-sensor/main.js"></script>

## How it works

The algorithm is embarrassingly simple for how well it works:

- Capture the time series of the accelerometer data (with a 1s half-life exponential decay),
- compute the PCA, and extract the first principal component. You now have a 1D time series
- Compute the FFT, and find the period with largest energy.

Smartphones: supercomputer processing power, tricorder sensors. Amazing stuff.