Signal Processing

To extract information from signals is an important part of our daily life. The capability to extract information using electronic devices is the capacity that signal processing provides. This very general knowledge area interacts with many disciplines such as image processing, speech recognition, neural processing, telecommunication theory, navigation systems and control systems.

Signal Processing knowledge is an important building block for much of modern technology in almost all areas of life. This means that the research a single group can contribute into is by necessity focused.

The WATRI Signal Processing Laboratory's main research concern is in the areas of:
          1. Acoustic signal processing
              a. Microphone arrays
              b. Echo cancellation
              c. Speech enhancement
          2. Optimisation in Signal Processing
              a. Filter bank design
              b. Optimal filter design
              c. Signal waveform set design
          3. Signal Processing Communications
              a. Pre-compensation schemes
              b. Antenna arrays and broadband arrays
              c. Equalisers

The common framework for the main part of the research is multi-rate signal processing which is gaining more and more importance in signal processing applications such as echo cancellation, microphone arrays, speech enhancement and equalisation. The research in the group is aimed at obtaining high performance and high efficiency signal processing designs by developing appropriate problem formulations based on solid mathematical foundations, in such a way that powerful optimisation techniques can be applied as well as a thorough understanding of electronic implementation constraints. The result will be that the developed multi-rate systems will have advantages in low complexity and low power hardware implementations without significant compromise in performance in each application.

Microphone arrays are used in numerous applications such as hands-free telephony, hearing aids, lap tops, audio conferencing etc. The theory of microphone arrays nest on theory developed for sonar arrays and antenna arrays but it has developed as an area of its own right due to the special requirements of microphone arrays. Further the applications where microphone arrays are used are typical low cost applications. Consequently the processing schemes used must allow for a large span of parameter variations.

Microphone array used for a hands-free application in a car.

Microphone array processing involves three main operations:
          - Localisation;
          - Spatial constraining;
          - Adaptive filtering.

The main difficulty with practical use of microphone arrays lies in the capability to localise a desired source and prevent the processing from suppressing the source of interest. Traditionally, this is achieved by adding constraints to the adaptive filters but experience have shown that this does not provide a satisfactory solution. Thus other forms of soft constraining of the solution have been more successful. Localisation is an important part to provide the soft constraint of the solution and also provide tracking of the user.

Current research lies in methods to improve localisation, speech quality and real time implementations. Blind signal separation methods are promising techniques and some initial work has also been done using blind signal separation techniques especially using ICA, Independent Component Analysis. The processing is performed using subband techniques to allow the use of more complex adaptive schemes while not making the computational load excessive.

Signal processing of acoustic signals has a numerous applications in sonar, medicine, audio, active noise control etc. Also problems that are related to hands-free use of communication terminals are essential. With terminals are meant home and office telephones, mobile phones, personal computers, video conferences and so on.

The research focuses on developing systems for remote sensing of speech signals in telecommunication applications, in particular hands-free telecommunication problems. The microphone is localised at a far distance from the user and the loud speaker should have a sound level that would allow you to convey a non-strained conversation. The microphone signal will pick up the desired speech signal, which has been distorted by the room response as well as background noise. Further by using hands-free communication there is a necessity to handle full duplex communication. The capability to use the equipment for speech recognition purpose is also very vital for the usefulness of the terminal. The techniques developed are also useful for general speech enhancement purposes.

The work is divided into three areas of research:
          - Microphone arrays, which includes tracking of speakers, echo suppression, reverberation suppression and
            background noise suppression;
          - Echo cancellation, suppression of acoustic echoes, double talk handling and post processing;
          - Speech enhancement, especially work on spectral subtraction and derivate methods.