A smart music player for real-time music-based feedback on tibial shock

A peak detection algorithm repetitively detected the magnitude and timing of tibial shock in each leg¹⁵. A custom-built JAVA program operated on the backpack system and detected a peak every time the axial acceleration exceeded 3 g with no higher axial acceleration value measured in the next 375 ms. This simple algorithm was based on a peak detection algorithm taken from a previous gait retraining study¹⁰. The magnitudes and timings were transmitted in real-time through Open Sound Control to a MAX/MSP patch that was built with the intention of providing music-based biofeedback in real-time²⁴. Real-time in this context means with negligible delay. For instance, when a new magnitude of tibial shock was detected, the auditory manipulations were executed in the same stride cycle.

The real-time, continuous, auditory biofeedback consisted of commercially available music tracks with superimposed pink noise of variable loudness (Fig. 2). The loudness of the noise depended on the momentary level of tibial shock of the leg that experienced the greatest mean shock in the baseline measurement. The five last values of that leg’s tibial shock were averaged through a 5-point moving average to account for inherent step-to-step variability in tibial shock¹⁰. That momentary level of tibial shock was mapped using an empirically validated fitting to obtain a distinct level of noise loudness²⁴. Six discrete loudness levels (0, 20, 40, 60, 80, 100% of noise) were created for good discretization (supplementary audio, fragment ¹)²⁴, thereby accounting for inter-subject differences in the decoding accuracies⁴². The loudness levels were calculated as a percentage of the root-mean-square amplitude level. So, the upper limit of 100% corresponded to noise with the same amplitude as the root-mean-square amplitude level of the music. Shock values below the target resulted in music only, meaning without pink noise (0% of noise). The target of minus ~ 50% of the baseline tibial shock was taken from previous gait retraining studies^7,8,10,17.

Schematic representation of the biofeedback system’s main components for continuous biofeedback on tibial shock (axial peak tibial acceleration). An interaction loop of the smart music player that provided the auditory biofeedback in real-time and that continuously accounted for (in)voluntary alterations in the running cadence by aligning the tempo (beats per minute) of the music to the cadence (steps per minute) of the runner. The red horizontal line indicates the baseline tibial shock. The five most recent values of tibial shock are averaged and mapped to a discretized level of noise loudness, which is added to the music playing.

The running cadence was derived from the timings of the tibial shocks detected during running. We intended to repetitively align the tempo of the music (i.e., the beats per minute) to the running cadence (i.e., the steps per minute) in the biofeedback condition^25,43. This real-time synchronization prevents the runner from adjusting his or her cadence to the tempo of the music and is based on the idea that interaction with music is empowering^27,28. Music of a preferred genre (pop, rock, electronic dance, swing, world) was chosen by the participant. A music database consisting of seventy-seven tracks with a clear beat in the tempo range of running at sub-maximal speed was created (supplementary information file, supplement ⁴). Songs with the right tempo were selected by a smart music player that instantaneously and continuously adjusted for a change in the running cadence. Music tempi were manipulated up to ± 4% of the steps per minute without pitch shift²⁴ (supplementary audio, fragment ²). When a change in steps per minute exceeded this tempo shift for eight seconds, another song started playing at a tempo that more closely resembled the altered running cadence. An illustrative audio fragment of a change in a music track was supplemented (supplementary audio, fragment ³). The momentary ratio of the music-to-motion alignment is described by the ratio of the running cadence (steps per minute) to the tempo of the music (beats per minute). The ratio should be close to 1 when the beats per minute of the music are aligned with the steps per minute of the runner.