Sound Analysis with Dynamic Impacts

Abstract

The impact analysis helps identify potential issues or provides a professional diagnosis, treatment, and management strategies. The first part of the research explored the interaction between spheres (tennis ball, ping-pong ball, pickleball, and lacrosse ball) and solid surfaces (wood, aluminum, and concrete surface) through impact experiments designed to collect sound signals and motion data. By varying experimental parameters, the study investigated how different factors, such as spheres and contact surfaces, affect impact dynamics. The audio feature analysis included dominant frequency, short-time pitch, and spectral entropy. The study examined the correlation between the kinematic coefficient of restitution (COR) and sound features. The dominant frequency had similar tendencies with COR when balls (excluding lacrosse balls) impacted different surfaces. The mean of short-time pitch and COR had similar tendencies when balls (excluding tennis and ping-pong balls) impacted different surfaces. The mean of spectral entropy had similar tendencies with COR when balls (excluding tennis balls and lacrosse balls) impacted different surfaces (excluding aluminum surfaces). The maximum sound pressure level and COR mostly had no similar tendencies for the collision. The deep learning method provided a convenient way to accurately distinguish the influence of various spheres and solid surfaces on impact. The second part of the research examined the footfalls of 25 horses trotted on a concrete surface. Sounds of footfalls were recorded, analyzed, and compared. At the same time, gait was evaluated to determine whether or not horses were lame using an inertial, sensor-based motion-analysis system. The study compared audio using audio features analysis and non-linear analysis methods. Statistical tests were applied to the characteristic audio signal parameters collected from sound and lame horses. The mean of short-time pitch results showed statistically significant differences in the audio signal features between sound and lame horses. Audio features recorded for horses while trotting might be a means of identifying lameness and potentially identifying the painful limb causing the lameness. The approximate entropy and largest Lyapunov exponent results showed significant differences in the audio signals between sound and lame horses. Lame horses exhibited lower values of approximate entropy, indicating more disordered and less predictable signal patterns, and higher largest Lyapunov exponent values, indicating less stability and regularity in their signals.