Home


Research

Grants

Publications

Courses

Vitae

Theses



 
 
 
 
 
 
 
 
Contact

Email
Research

     Research interests:

    Speech Processing

    There has been a radical change in the way people communicate today. There exist many options for communicating or transmitting information over the wired network, including voice mail, beeper services, ISDN video, FAX, audio and video teleconferencing, etc. One of the main technologies that made this possible, is the speech processing technology, which includes speech compression (compression of the signal for transmission though limited-bandwidth channels), and speech recognition (recognition of voice input by the machine). Dr. Loizou has been involved in several projects related to speech compression, speech recognition, and speech enhancement. For more information about our speech processing research visit our Speech Processing Lab 
  • Speech enhancement

    • Speech enhancement techniques are used to remove noise from the speech signal or enhance the speech quality.

         Current Projects:

o       Development of MMSE algorithms for speech enhancement.

o       Development of noise estimation algorithms suitable for highly non-stationary noisy conditions

o       Development of subspace algorithms for non-stationary environments

o       Active noise control for fMRI scanner rooms
                       

[Research supported in part by NIH and done  in collaboration with University of Texas Southwestern Medical center (UTSW)]


 

  • Speech Compression

    • Speech compression (or coding) is needed for efficient transmission and storage of the speech signal. For transmission applications over the telephone network, the goal of speech compression is to transmit the signal over limited-bandwidth channels while maintaining voice quality and intelligibility. 

    Past Projects:

    •  Real-time implementation of speech compression algorithms, including VSELP which has been adopted as the North American digital cellular standard, and GSM, the European speech coding standard.
    • Design and implementation of vector quantization algorithms for speech coding 

    Top 
     

  • Speech Recognition

    • The goal of automatic speech recognition (i.e., recognition of speech by machines) is to interact with the machines via voice commands. Such a technology could be very valuable in myriads of applications, such as voice dialing of telephone numbers, automation of database access via voice input, voice dictation systems, etc. 


    Past Projects:

    • Development of preprocessing algorithms for digit recognition. Preprocessing algorithms were run real-time on a fixed-point processor. 
    • Analysis of neural network algorithms for automatic vowel classification 
    • Development of alphabet recognition algorithms for name retrieval applications using context-dependent Hidden Markov Models. 


    Top 

    Cochlear Implants

    Several million Americans today have profound hearing loss, and for years they had to rely on conventional hearing aids. Although hearing aids have been found to benefit hearing impaired individuals who suffer moderate deafness, they have not been found to benefit individuals with severe (sensorineural) deafness. Cochlear implants are now established as a new option for individuals with profound hearing impairments. Many of these individuals, who are implanted with cochlear prosthesis, are able to understand some speech without lip-reading. Several speech-sound processing techniques have been developed over the years that improved the benefits derived from the cochlear implant. Moderate levels of speech understanding can now be achieved with current speech-processing techniques. Dr. Loizou's research focuses on the development of new speech-sound processing strategies which will help improve the levels of speech performance even further.  For more information about our cochlear implant research (supported by NIH) visit our Cochlear Implant Lab.

    Signal processing for cochlear implants

    o       Development of signal processing algorithms for cochlear implant processors

    o       Development of speech coding algorithms for music

    o       Development of noise reduction algorithms for cochlear implants

    o       Speech perception by cochlear implant patients

    o       Fixed-point implementation of speech processing algorithms on Motorola's DSP56002 and Texas Instrument’s TMS320C54x DSPs

    o       Optimization of signal processing algorithms on TMS320C6x DSP

    [Research supported by NIH and by Advanced Bionics Corporation]

     

    Top