Application of Wavelet Filtering to Analyze Acceleration-Time Curves of Horses Trotted on Different Surfaces

The objective of the present study was to illustrate the possibilities of wavelet filtering to denoise hoof-acceleration data measured on different surfaces. Further, three variables from the time-domain acceleration signal and two variables from the single-sided amplitude spectrum of the Fourier-transformed signal were statistically analyzed to detect the differences between the riding surfaces as well as to compare the results of the different variables. For data recording, six Warmblood horses were trotted on five different riding surfaces. The recorded acceleration data were smoothed by wavelet filtering on three levels by using the Haar wavelet as well as the fourth-order Daubechies wavelet. The hoof’s landing phase was extracted from the acceleration signal by a special algorithm and analyzed because of to its significance with regard to the injury risk of a horse. More accurate hoof’s landing phases could be detected by the algorithm in the first and second approximations of both wavelet-filtered signals than in the original signal. The dissected hoof’s landing phases were used to calculate the aforementioned variables. The statistical analysis showed similar results within the variables of the time-domain signal and within the variables of the single-sided amplitude spectrum. The least square means of the original signal and the first and second approximation of both wavelet-filtered signals were on a similar level. In conclusion, it was shown that wavelet filtering is a suitable method to denoise acceleration-time curves. Further, the results of the five variables showed that the insertion of time- and frequency-domain variables was significant in the analysis of the acceleration data.