# The Effect of Filtering on Signal Features of Equine sEMG Collected During Overground Locomotion in Basic Gaits

## TL;DR

This study shows how different filtering methods affect equine sEMG signals during basic gaits, highlighting the importance of standardized filtering for reliable comparisons in equine research.

## Contribution

The study identifies optimal filtering methods for equine sEMG signals to balance noise reduction and signal preservation.

## Key findings

- High-pass filtering at 40 Hz and bandpass filtering at 40–450 Hz significantly changed signal features while fully attenuating low-frequency noise.
- Other filtering methods resulted in partial noise attenuation and less consistent signal feature changes across gaits.
- Standardized filtering is recommended to ensure reliable comparisons in equine sEMG research.

## Abstract

In equine surface electromyography (sEMG), challenges related to the reliability and interpretability of data arise, among other factors, from methodological differences, including signal processing and analysis. The aim of this study is to demonstrate the filtering–induced changes in basic signal features in relation to the balance between signal loss and noise attenuation. Raw sEMG signals were collected from the quadriceps muscle of six horses during walk, trot, and canter and then filtered using eight filtering methods with varying cut–off frequencies (low–pass at 10 Hz, high–pass at 20 Hz and 40 Hz, and bandpass at 20–450 Hz, 40–450 Hz, 7–200 Hz, 15–500 Hz, and 30–500 Hz). For each signal variation, signal features—such as amplitude, root mean square (RMS), integrated electromyography (iEMG), median frequency (MF), and signal–to–noise ratio (SNR)—along with signal loss metrics and power spectral density (PSD), were calculated. High–pass filtering at 40 Hz and bandpass filtering at 40–450 Hz introduced significant filtering–induced changes in signal features while providing full attenuation of low–frequency noise contamination, with no observed differences in signal loss between these two methods. Other filtering methods led to only partial attenuation of low–frequency noise, resulting in lower signal loss and less consistent changes across gaits in signal features. Therefore, filtering–induced changes should be carefully considered when comparing signal features from studies using different filtering approaches. These findings may support cross-referencing in equine sEMG research related to training, rehabilitation programs, and the diagnosis of musculoskeletal diseases, and emphasize the importance of applying standardized filtering methods, particularly with a high–pass cut–off frequency set at 40 Hz.