PPubMed28 Jan 2011Cited 68×

Leg Motion Classification with Artificial Neural Networks Using Wavelet-Based Features of Gyroscope Signals

Birsel Ayrulu-Erdem, Billur Barshan

Abstract

We extract the informative features of gyroscope signals using the discrete wavelet transform (DWT) decomposition and provide them as input to multi-layer feed-forward artificial neural networks (ANNs) for leg motion classification. Since the DWT is based on correlating the analyzed signal with a prototype wavelet function, selection of the wavelet type can influence the performance of wavelet-based applications significantly. We also investigate the effect of selecting different wavelet families on classification accuracy and ANN complexity and provide a comparison between them. The maximum classification accuracy of 97.7% is achieved with the Daubechies wavelet of order 16 and the reverse bi-orthogonal (RBO) wavelet of order 3.1, both with similar ANN complexity. However, the RBO 3.1 wavelet is preferable because of its lower computational complexity in the DWT decomposition and reconstruction.

§ The Valyu brief

The paper, in five minutes

Read from the full text via Valyu Contents

Wavelet-based features from low-cost gyroscope signals enable ANN classification of eight leg motions with up to 97.7% accuracy and low network complexity.

What’s new

i.Extracted informative wavelet-based features from gyroscope signals for leg motion classification.
ii.Systematically evaluated the impact of different wavelet families and orders on classification accuracy and ANN complexity.
iii.Identified a 12‑element feature set (normalized variances at level 5) achieving 97.6% accuracy with low ANN complexity.
iv.Demonstrated the use of low‑cost leg‑mounted gyroscopes for potential physiotherapy and home‑based rehabilitation monitoring.

How they did it

Gyroscope data from two sensors on the human leg were segmented into 1,600‑sample windows and decomposed using the discrete wavelet transform with various wavelet families. Features such as normalized coefficients, energy distribution ratios, and statistical measures were extracted and fed into three‑layer feed‑forward ANNs trained with the Levenberg‑Marquardt algorithm. Classification performance was assessed on a test set comprising one‑third of the 768 recorded motion segments.

What they found

Maximum classification accuracy of 97.7% achieved with Daubechies wavelet order 16 and reverse bi‑orthogonal (RBO) wavelet order 3.1.
Average classification accuracy of 97.6% using the 12‑element normalized variance feature set with the RBO 3.1 wavelet.
Wavelet family averages: Coiflets 93.4%, RBO 93.3%, Symlets 90.1%, Daubechies 89.5% on test data.
Using normalized variances alone yielded 95.1% test accuracy at decomposition level 5 with ANN architecture 12:8:8.

Caveats

✕Experiments were conducted with data from a single subject, limiting assessment of inter‑subject variability.
✕Only eight predefined leg motions were studied, restricting generalization to other movements.
✕The approach relies on offline processing; real‑time implementation was not evaluated.
✕Classification depends solely on gyroscope data, without fusion of additional sensor modalities.

§ The Valyu brief

Reading the full paper and taking notes. This takes a few seconds…

§ Ask this paper

Ask a question about this paper

Valyu reads the full text and answers from what the paper actually says.

§ Related work in other archives

Leg Motion Classification with Artificial Neural Networks Using Wavelet-Based Features of Gyroscope Signals

The paper, in five minutes

Ask a question about this paper

Similar papers in all four archives

Similar papers in all four archives

The paper, in five minutes