Use this URL to cite or link to this record in EThOS:
Title: Advanced multimodal approach for non-tagged indoor human identification and tracking using smart floor and pyroelectric infrared sensors
Author: Al-Naimi, Ibrahim
ISNI:       0000 0004 2707 5416
Awarding Body: De Montfort University
Current Institution: De Montfort University
Date of Award: 2011
Availability of Full Text:
Access from EThOS:
Access from Institution:
Significant research efforts have been directed into smart home environments in the last decade creating abundant opportunities for the broader home services ecosystem to foster a wide range of innovative services. Research interest has been given on automatic identification and tracking of people within the home environment to support customised services such as care services for elderly and disadvantaged people to enable and prolong their independent living. Although various approaches have been proposed to tackle this problem, solutions still remain elusive due to various reasons (e.g. user acceptance). Literature reviews have indicated the need for an advanced non-tagged identification and tracking approach that is capable to provide the infrastructure support for realisation of context-aware services, satisfy users’ needs, and deal with the complexity of smart home environmental conditions. The aim of this study is to develop and implement an advanced approach that is capable to accurately detect, identify, and track people within opportune and calm home environment to be used as infrastructure for various application domains such as assisted living, healthcare, security and energy management. Accordingly, a novel multimodal approach for non-tagged human identification and tracking within home environment is proposed. The proposed approach combined floor pressure and PIR sensors through unique designed integration strategy aiming to merge the advantages of the two sensor types and overcome or minimise their weaknesses. The designed strategy enabled the PIR output signal pattern to afford explicit information indicating a person’s body surface area (size/shape). This information enhanced the identification accuracy, facilitated the custom designed smart floor, and reduced the overall cost. The conceptual framework of the proposed approach/strategy encompassed two key stages, hardware system design and implementation, and data processing. The hardware system design included the custom designed PIR and smart floor units. A test bed was designed and implemented for supporting the research studies, including proof of concept, concept demonstration, experimental and test cases studies. Data processing system has divided into different stages to accomplish the identification and tracking goals. First, the interested patterns were segmented and generated with threshold edge detection method and advanced pattern generation algorithm respectively. Second, limited set of features were extracted and selected from each pattern including ground reaction force GRF, gait, and body size/shape (PIR) features. Third, these features were merged at different fusion level, namely, feature-level and decision-level to provide comprehensive description about the person’s identity. Fourth, MLPNN multiclass classifier was adopted to process the feature vectors and recognise the person’s identity. Finally, the footstep patterns were tracked using weighted centroid tracking technique, in addition to MLPNN classifier to handle the footsteps association problems. Four test cases were designed and carried out to demonstrate, test, and evaluate the feasibility and effectiveness of the proposed non-tagged identification and tracking strategies/approach. The assessment outcomes have shown the potential of the proposed multimodal approach as an advanced strategy for implementation of an indoor non-tagged human identification and tracking system and to be used as infrastructure for supporting the delivery of various types of smart services within the smart home environments. In summary, the proposed multimodal approach has the potential to: (1) Identify up to 5 persons successfully with minimum 98.8% correct classification rate without tag, (2) detect, locate, and track multiple persons successfully without tag and the location error no more than 11.76 cm, approximately 1.5 times better in accuracy than the original set target (i.e. 30 cm), and (3) able to handle various tracking difficulties and solve 97.5% of data association problems.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available
Keywords: multimodal tracking approach ; indoor tracking and identification ; smart floor ; pyroelectric sensors