Indoor Localisation Using Received Signal Strength
Name
Henry Teigar
Abstract
Indoor localisation of people has gained a lot of interest during the last decade. Different approaches have been proposed and tested in various environments. This thesis tries to predict a person’s location in the SPHERE testing house. SPHERE is a project with an aim to use sensor technology for healthcare, such as early diagnosis of different illnesses by monitoring person’s activity in their homes. Accurate localisation of the person can provide useful information for this purpose. We use the received signal strength indicator (RSSI) values between the receivers with fixed positions and one mobile node to perform the localisation. For this we use two machine learning methods: hidden Markov models (HMMs) and k-nearest neighbors algorithm (k-NN). A detailed
description of the implementation process of both models used on the SPEHRE dataset is also given. Finally, we provide the results and the comparison of both approaches.
We found that after feature pre-processing, the k-NN performed surprisingly well by achieving room-level accuracies around 90%. The initial performance of the HMM was found to be similar to k-NN’s but with our modifications to the HMM, we finally achieved accuracies up to 96%.
description of the implementation process of both models used on the SPEHRE dataset is also given. Finally, we provide the results and the comparison of both approaches.
We found that after feature pre-processing, the k-NN performed surprisingly well by achieving room-level accuracies around 90%. The initial performance of the HMM was found to be similar to k-NN’s but with our modifications to the HMM, we finally achieved accuracies up to 96%.
Graduation Thesis language
English
Graduation Thesis type
Bachelor - Computer Science
Supervisor(s)
Meelis Kull
Defence year
2018