Biologically Inspired Speaker Verification - Voice biometrics is one of the least developed of all the biometric technologies. Environmental noise, intra- speaker feature variability and platform dependent speech-processing differences are all factors that substantially limit the performance of existing speaker verification algorithms (GMM, SVMs etc.). Neural network based methods have the potential to automatically address these factors as part of their training process, and novel supervised, unsupervised and hybrid neural architectures have been developed within the speech-enabled systems research group here at NTU. Current research is investigating the use of Spiking Neural Network systems, based on models of the human auditory system, as a more biologically plausible speaker verification method. Temporal Issues in designing Fuzzy Systems - During past years, fuzzy methodologies have emerged as one of the most suitable and efficient methods for designing and developing complex systems in environments characterized by high level of uncertainty and imprecision. Nowadays, this methodology is used to model systems in several applications domains which range from industrial machineries to financial decisions support systems. Nevertheless, in spite of the usefulness of fuzzy logic, one of its drawbacks comes from the lack of the temporal concept that is crucial in many systems characterized from a discontinuous nonlinear behaviour. In particular, in its standard vision, fuzzy control is not able to represent Variable-Structure systems, i.e., systems that change their configuration (knowledge base) or their behaviour (rule base) over time. To overcome these drawbacks, this research extends fuzzy systems idea by considering a theory from formal languages: timed automata. This novel synergic approach achieves a twofold advantage by representing a system in qualitative and linguistic way and introducing a novel switching control concept able to maximize system's performances and robustness. This approach can be applied to different application scenarios as computer networking security, urban garbage collection and so on. Recurrent Neural Networks for Modelling Temporal Phenomena - Much of the world around us is dynamic and temporal in nature – daily weather patterns, night/day time transitions, the changing seasons, our periodic heart beats and more generally, daily living patterns. Our world is full of time-dependent processes and almost any real system has some notion of temporal state and transition, from calling a lift, using a vending machine to even switching on a light. The human use of natural language to communicate is a classic temporal problem whereby structure and meaning is encoded within linear strings of inter- connected words. Temporal based Neural Networks can be also integrated in complex computer vision systems in order to identify some hard and dynamic patterns occurring in real environment such as human activities and behaviours. Middleware, Domain Specific Languages and Trust – Getting heterogeneous devices and services to communicate and understand each other is problematic within pervasive computing environments. An active area of research is through the use of light-weight service composition middleware's built to allow communication and coordination to allow the re-configuration and identification of services. This can be coupled with domain specific languages (DSL) that enable services to be easily created and to hide the complexity of the environment. In past work, DSL's were used to program these pervasive environments, while more recently, DSL's are being investigated in the control of power consumption within smart homes. With a large number of services available within a pervasive environment, trust becomes an issue and work in to rule based provenance trust systems are currently being investigated.
CIA_Activities Classification and Abnormality Detection
Infrared thermal imaging of buildings with different insulations
Frequency distribution of the lounge temperature for a given outside ambient temperature
Made with FlippingBook flipbook maker