Acoustic gait analysis

According to the World Health Organization, the number of Parkinson's cases worldwide has doubled. As the second most common neurodegenerative disease, Parkinson's affects the gait of those affected. Other diseases, such as muscle disorders, also have a negative impact on gait. In view of demographic change, providing medical care for an ageing population poses challenges for many European countries, including Germany. Early diagnosis is crucial, and gait analysis can play an important role in this process by identifying and evaluating changes in gait.

An innovative concept – the acoustic shoe
At the Fraunhofer IDMT, an innovative project for mobile acoustic gait analysis has been started that is based on the novel concept of the acoustic shoe. The prototype consists of a flexible fabric material and a non-slip sole that allows stability, comfort and unrestricted wear over a longer period of time. A highly sensitive MEMS microphone captures the acoustic signals of walking, while strategically placed pressure sensors record ground contact. These pressure sensors are used to accurately detect and differentiate the stance and swing phases of the gait. They create annotations for the acoustic data sets, which are crucial for extracting the relevant audio segments of each gait phase. The aim is to use this data to better analyse and classify the acoustic signals during development.

The collected data is then processed by a single-board computer that serves as a central processing unit. This system thus enables the parallel acquisition and synchronisation of audio and pressure sensor data. Tests were carried out to determine whether classifying accuracy could be improved by editing the audio data based on the ground contact times recorded by the pressure sensors. However, it was found that removing noise by editing the data to only include the stance phases did not significantly improve the classification results. The long-term goal is to implement the trained algorithms on a compact and energy-efficient edge platform. This should work in a mobile application or by means of wireless data transmission, e.g. in a stationary environment. The aim is to replace the pressure sensors and to use only the acoustic data from one or more sound receivers to recognise and assess gait patterns.

Feasibility study: analyzing gaits
In a feasibility study with 39 subjects, the acoustic shoe was used to analyze three exemplary gaits: normal gait and two pathological gaits – myopathic gait and Parkinson's gait. The subjects simulated the different gaits under instruction. Despite the limited sample size, the study provides a valuable database for an initial assessment of the acoustic characteristics and allows trends and patterns to be identified that are relevant for further studies.

In addition to conventional audio signal processing methods, advanced algorithms such as k-Nearest Neighbours (k-NN) and K-Means Clustering were used for the data analysis. These methods were used to identify and classify characteristic features of the gait. The analysis already showed a classification accuracy of up to 71 percent, which is well above the random level of 33.3 percent. It is particularly noteworthy that the acoustic shoe was able to detect not only normal gaits, but also simulated pathological gaits with a high degree of accuracy.

Potential for research and application 
The acoustic shoe marks a significant advance in the development of wearable gait analysis technology. The successful feasibility study opens up exciting prospects for further research and development, particularly in medical diagnostics, where the shoe can serve as a valuable tool for monitoring and analyzing gait disorders. Furthermore, this technology also offers potential for applications in rehabilitation, self-monitoring and as part of ‘smart shoes’ for athletes and health-conscious end users.