Acoustic Fall Detection

Project description

Enabling older persons to continue living independently at home has many benefits, individually, for families and for society at large. One of the challenges to facilitating this in the face of age-related frailty and/or co-morbidities, such as diabetes, dementia or mobility impairing illnesses is ensuring that personal safety can be assured to a satisfactory degree, and that should the person fall, that this can be detected and responded to rapidly. Various methods have been proposed for detecting falls in the elderly. In this project, the goal is to explore the potential for passive and active acoustic arrays for the localisation and classification of sound sources. Our hypothesis is that human vocalisations that originate from below a given height (altitude) in a room are likely to be the result of a person having fallen. To test this hypothesis we require the construction of a low-cost and accurate acoustic array that has sufficient angular and range resolution to allow the discrimination between sound sources based on their altitude in the room. We desire to do this using a thin-line array, effectively a string of acoustic sensors (microphones), plus one or more transducers to also allow active operation. A key challenge in the construction and operation of such thin-line arrays is ensuring that the position of each microphone in the array, so that beam forming and other localisation methods can be employed. This includes determining the location of each element during assembly, as well as variation in relative position due to bending of the array. Further, for many such systems, they are designed such that assumptions about their straightness must be made, which almost always introduces errors, and also disregards the information that a non-straight array can provide compared with a straight array, precisely because the sensors cannot all be placed in a single vector or plane. We propose to address these problems by creating a modular system of small acoustic sensor and transducer boards that are use acoustic methods to determine their relative position and orientation to one another, thus allowing for real-time continuous calibration., regardless of the 3D shape that the array may take. For the proposed aged-care context this has the further benefit of allowing the array to be deployed in any shape, and without requiring it to be either rigid or straight. This year we are looking at two sub-projects: 1. Electronic Engineering: Select appropriate microphones, transducers and other components, and design, fabrication and test of an initial version of the modular board. 2. Mathematics, Statistics, Computer Science or Engineering: Analyse the theoretical performance of an array of such modules, including determining the likely positional and orientation accuracy that can be obtained, and then base on that, the likely spatial resolution of the resulting array using well-known sonar array equations, experimental simulation, or a combination of the two. Note that the mathematics for this is not particularly challenging, consisting of basic algebra and some trigonometry. The signal processing can be performed entirely in the time domain.

Co-supervisors

Paul Gardner-Stephen https://www.flinders.edu.au/people/paul.gardner-stephen

Note: You need to register interest in projects from different supervisors (not a number of projects with the one supervisor).
You must also contact each supervisor directly to discuss both the project details and your suitability to undertake the project.