Abstract
Acoustic localisation technology promoted by modern engineering is of significant meaning commercially, militarily and medically. Commercially, Human-Computer Interfaces (HCI) powered by the acoustic localisation technology are attractive to consumers. Militarily, modern Sound Navigation and Ranging (SONAR) systems are essential equipment for the Navy and the acoustic localisation technology is one of the core technologies in SONAR systems. Medically, acoustic localisation technology is used to locate and smash stones inside human bodies. Acoustic localisation technology has a series of applications in various fields. However, the indoor localisation is an exception.The localisation demands of individuals in indoor environments are rising. For example, the Virtual Reality (VR) technology allows users to explore virtual worlds, but the VR technology also brings challenges to human-computer interaction. Traditional Time Difference of Arrival-based electromagnetic localisation technologies and touchtone HCIs cannot provide users with immersive user experiences. However, HCIs powered by the acoustic localisation technology can locate users passively and such HCIs may improve the user experience by removing mobile restrictions of touchtone HCIs and wearable devices.
At present, two approaches are applicable to achieve acoustic localisation. The first approach is the time difference localisation. The time difference acoustic localisation technology locates an acoustic source according to the time differences measured by sensors located at multiple locations. The Time Difference of Arrival (TDOA) is the most representative time difference localisation methodology. However, TDOA-based localisation technologies are inaccurate in short-range scenarios due to the interferences caused by the multipath effect. The second approach is the Template Pattern Matching (TPM) localisation. It is a similarity-based matching methodology developed for string examination in computer science originally. Regarding localisation, TPM-based acoustic localisation technologies locate acoustic sources by matching the acoustic signals of the acoustic sources with the template acoustic signals. The template acoustic signals are previously collected from pre-defined locations. In template signal matching, the template acoustic signal which has the highest similarity to the input acoustic signal is identified as the matched signal thus the acoustic source is located at the pre-defined location of the matched template signal. The TPM-based acoustic localisation system has a reasonable system cost since it does not rely on the deployment of multiple sensors and system synchronisation in comparison to the TDOA-based localisation system. But TPM-based localisation technologies need adequate data and exact matching algorithms to ensure accurate matching results. A representative application of TPM is the passive SONAR system on submarines. This project aims to achieve three-dimensional short-range acoustic source localisation with the TPM approach.
The short-range acoustic localisation technology has been applied to two-dimensional human-computer interaction but has not been applied to three-dimensional human-computer interaction. Since the short-range localisation technology is a key enabling technology, current research on three-dimensional human-computer interaction tends to realise three-dimensional human-computer interaction with the optimised TDOA-based electromagnetic localisation technology. However, specific problems such as the deployment of sensors and the high system cost are exposed in the implementations. These disadvantages restrict the applicability of the short-range localisation technology; thus, an acoustic localisation technology is developed in this project to overcome the disadvantages of the TDOA-based electromagnetic localisation technology and enhance the applicability of the short-range localisation technology.
Acoustic waves can be generated actively by speaking, knocking and tapping therefore acoustic sources are ideal signal sources for passive localisation. In this thesis, a passive pattern matching-based acoustic localisation technology - Location Template-based Positioning Model (LTPM) is successfully designed, implemented and tested in terms of three-dimensional human-computer interaction. The proposed technology determines the coordinates of acoustic sources by matching input acoustic signals with pre-collected template signals according to 43 acoustic features and 17 signal images. The test results indicate that LTPM has achieved a three-dimensional accuracy of 173 mm in 95% of the location estimates in an indoor environment.
LTPM successfully utilises the acoustic multipath effect to achieve three-dimensional acoustic localisation and LTPM does not rely on the deployment of sensor arrays, measurement of time difference of arrival, path optimisations and signal filtering in comparison to the TDOA-based positioning system. Meanwhile, the LTPM-based positioning system has robust environmental adaptability. Correspondingly, the positioning performance of LTPM depends on the data volume, the integrated features, and the matching accuracy of the algorithm.
In this thesis, LTPM is applied to two-dimensional surfaces and an indoor three-dimensional space for localisation tests. The two-dimensional LTPM-based localisation system has achieved a two-dimensional accuracy of 30 mm in 80% of the location estimates. While the three-dimensional LTPM-based localisation system has achieved a three-dimensional accuracy of 173 mm in 98% of the location estimates. These results imply that three-dimensional acoustic localisation in indoor environments is feasible, and LTPM has the potential to be the enabling technology for three-dimensional human-computer interfaces.
| Date of Award | Aug 2023 |
|---|---|
| Original language | English |
| Awarding Institution |
|
| Supervisor | Ming Yang (Supervisor), Xiang Fei (Supervisor) & Kuo-Ming Chao (Supervisor) |