# Learning-Based Speed Estimation from Accelerometer-Only Inertial Sensing

arXiv: https://arxiv.org/abs/2401.07468 · HTML: https://mlsanity.com/paper/2401.07468/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 8 of its 20 references that are in the map.
- First posted on arXiv 15 Jan 2024.

## Builds on (most cited first, 8 of 8)

- [Deep Residual Learning for Image Recognition](https://mlsanity.com/paper/1512.03385/) (2015)
- [Adam: A Method for Stochastic Optimization](https://mlsanity.com/paper/1412.6980/) (2014)
- [WaveNet: A Generative Model for Raw Audio](https://mlsanity.com/paper/1609.03499/) (2016)
- [AI-IMU Dead-Reckoning](https://mlsanity.com/paper/1904.06064/) (2019)
- [Learning Vehicle Trajectory Uncertainty](https://mlsanity.com/paper/2206.04409/) (2022)
- [Learning Car Speed Using Inertial Sensors for Dead Reckoning Navigation](https://mlsanity.com/paper/2205.07883/) (2022)
- [Deep Learning for Inertial Sensor Alignment](https://mlsanity.com/paper/2212.11120/) (2022)
- [Learning Position From Vehicle Vibration Using an Inertial Measurement Unit](https://mlsanity.com/paper/2303.03942/) (2023)

## Abstract (arXiv)

The proposed model, CarSpeedNet, estimates scalar vehicle speed from a window of three-axis smartphone acceleration, without gyroscope, wheel-odometry, vehicle-bus, or positioning input at inference. The reported experiment comprises 13.2 hours of on-road driving. Beyond the network comparison, a finite-context analysis treats window length as part of the sensing problem. For nested histories, the minimum Bayes mean-square error is non-increasing with context; a complementary cue-coverage relation links the same window to the amount of speed-dependent vibration presented to the network and to the observation horizon. At a 1-s input, CarSpeedNet is compared with five temporal-network alternatives; the effect of temporal context is then measured over six window lengths. On the 0.5-hour holdout, a 4-s window yielded a root-mean-square error (RMSE) of 1.8 m/s and a mean absolute error (MAE) of 0.72 m/s, compared with 2.9 and 1.3 m/s at 1 s. The 178,169-parameter network requires about 0.68 MiB for 32-bit weights.