Jinsoo Jeong

Doctor of Philosophy, (Information Engineering)
Study Completed: 2007
College of Sciences

Citation

Thesis Title
A Kepstrum Approach to a Real-Time Speech Enhancement.

Read article at Massey Research Online: MRO icon

Mr Jeong proposed a robust method for improved performance of real-time speech enhancement and noise cancellation in a real reverberant environment. Hepresented the kepstrum approach, based on a fundamental theory of kepstrum analysis, and its application to a speech enhancement method. Kepstrum analysis was applied to system identification of unknown acoustic transfer functions between two microphones, with a modified application technique from conventional adaptive noise cancelling and beamforming. This kepstrum method provided a mathematical construct with FFT based processing and was independent of acoustic path model order. The research has shown that the front-end application of the kepstrum method to speech enhancement methods provides an improved performance in speech enhancement and noise cancellation with several favourable effects.

Supervisors
Dr Tom Moir
Dr Mohammad Abdurrashid