Analysis of human activities and animal behaviours based on computational intelligence



Alzu'Bi, Hamzah
Analysis of human activities and animal behaviours based on computational intelligence. [Unspecified]

[img] Text
AlZu'biHam_Nov2015_2037039.pdf
Access to this file is embargoed until Unspecified.
Available under License Creative Commons Attribution.

Download (36MB)
[img] Text
AlZu'biHam_Nov2015_2037039.pdf
Available under License Creative Commons Attribution.

Download (36MB)

Abstract

The study of behaviour is vital for animal welfare assessment in animal husbandry systems, exploring mechanisms of underlying diverse forms of behaviours and animal physiological and ecological interaction. Understanding animal behaviour is used in a systematic way to unlock and explore underlying functionalities of the brain which is one of the biggest challenges to science. This thesis introduces four novel applications for computational intelligence in human and animal behaviours. The four applications are: horse transport stress prediction system, human activity recognition, fish behaviour tracking and detection, and intelligent interactive fish feeding system. In the first application of human gait recognition, a practical, accurate and novel supervised learning system is proposed to recognize human activities. The proposed system uses single accelerometer device which makes the system practical to use and capable of being integrated with many commercially available devices. This work proposes highly accurate and practical human gait recognition system. In the second application of horse transport, a novel system is proposed to predict horse stress episodes during transport which enables a potential solution of horse stress by interfering at a suitable time. Dynamic nonlinear neural network is trained to predict horse stress time series given travel route and driving style time series. Horse transport is one of the most routinely stressful procedures in equine industry. In the third application of horse transport, a novel system for automatic fish tracking and behaviour recognition system is proposed. Fish are the second most popular experimental model behind mice in pharmaceuticals and biological research. Fish anxious behaviour could confound experiment outcomes. Fish behaviour could also be affected by invasive or non-invasive experiments in addition to other possible causes of distress. The proposed system consists of 3d real-time fish tracking, behaviour quantifying and recognition algorithms. Fish behaviour is estimated through fish swimming patterns. The system showed high accuracy recognition of fish behaviour in experiment where fish were exposed to a variety of external stimuli. In the fourth application of horse transport, an innovative smart fish feeding system is proposed. The fourth application of computational intelligence techniques addresses one of the major challenges in the fastest growing food sector industry worldwide, aquaculture industry. Most conventional fish feeding techniques are inefficient, cause environmental damage and fish losses, raise concerns regarding fish welfare and lead to non-uniform fish growth. Addressing these problems is a necessity for this industry to continue its growth. The novel feeding system is built based on fish behaviour which recognises, and assesses fish behaviours and interacts with fish to optimise the feeding process. Fish showed quick adaptation to this novel low-cost feeding system which proves the feasibly of implementing this system. The proposed system is expected to reduce food competition and environmental impact because of its responsive nature. Through novel applications of computational intelligence, this thesis has provided successful solutions for human and animal behaviour analysis research problems.

Item Type: Unspecified
Additional Information: Date: 2015-11 (completed)
Subjects: S Agriculture > S Agriculture (General)
S Agriculture > SF Animal culture
T Technology > TK Electrical engineering. Electronics Nuclear engineering
Depositing User: Symplectic Admin
Date Deposited: 18 Jan 2016 11:00
Last Modified: 03 Jan 2021 01:18
URI: https://livrepository.liverpool.ac.uk/id/eprint/2037039