Projects that have been completed by the digital media processing and biometrics team are shown here.
This was an EPSRC-funded Network in Biometrics, with partners from University of Kent, University of Surrey, University of Southampton, University of Sheffield and University of Nottingham.
Its aim was to bring together the key academic researchers in the UK biometrics field and to develop and strengthen the interface with the industrial base in order to maximise the effectiveness of biometrics in crime prevention and detection.
A factor important to the success of a variety of ubiquitous computing applications is that of identification of users so that each individual is provided with an appropriate level of authorisation.
An appropriate approach to identification in such applications is that of voice biometrics. The work was mainly concerned with the development of a system for the extraction of robust speech features in uncontrolled environments.
Such a feature extraction engine can be used to present a standardised parametric model for the representation of voice characteristics that a subsequent classification system may utilise to achieve signature matching. The project involved the development of an integrated noise cancellation-feature extraction system and a demonstrator for a user classification scenario.
The project was funded under the Next Wave Technologies and Markets programme, and was conducted in collaboration with Fulcrum Voice Technologies (FVT) and in association with the DTI Virtual Research Centre at York University.
This was an EPSRC funded project conducted in collaboration with industry. It involved the development of effective methods for subtitling live television programmes using speech and speaker recognition.
The work included investigations into multiple-domain language modelling, methods for reducing speech recognition errors, and environmental robustness. Another important area of investigation was the development of methods for automatic colour coding of subtitles based on speaker separation and tracking.
The technique developed for this purpose was capable of segregating speakers without prior knowledge of the identities of the speakers.
This was a collaborative project, and its main focus was the reduction of in-band alarm noise in telephony and radio communications for emergency services. The research conducted in the field led to the development of a novel method for the minimisation of alarm noise in speech.
The main objective of this project was to investigate effective methods for the recognition of people over the Internet based on their voice and facial characteristics.
The motivation behind this research was the need for a reliable means of identification to complement/substitute the traditional identity verification approaches deployed for logical access to services over the Internet.
The project was undertaken in collaboration with industrial partners and in association with COST Action 275.
Exploration and map-building of an unknown environment is one of the main issues in mobile robotics due to its wide range of practical applications, including search and rescue, hazardous material handling, military actions, planetary exploration, path planning and devastated area exploration.
The purpose of this research project was to investigate the field of exploration and map building with cooperative mobile robots. A frontier-based technique and a line-of-sight technique have been employed with the intention to reduce the exploration time required to build a grid-based map of an unknown environment.
MPEG variable bit rate (VBR) data sources experience unpredictability, long delay, and excessive loss, due to sudden variations in bit rate. Therefore, it is practically impossible to transmit MPEG-2 VBR video sources over a Bluetooth channel with a limited transmission speed and considerable wireless interferences without loss of data or image quality degradation.
In this work, a traffic-shaping buffer is introduced before the host controller interface of the Bluetooth protocol stack to obstruct excessive MPEG VBR data sources during the peak transmission periods.
An integrated neuro-fuzzy controller monitors and reduces the burstiness of the traffic-shaper output rate to facilitate the MPEG VBR video sources to conform to the token-bucket contract before entering the Bluetooth channel.