(Dr. Jerry Ginsberg, advisor)
"A Methodology for Studying Distortion of Speech Resulting From Propagation Through a Layered Wall"
An algorithm previously developed to study the distortion of a planar stress wave as it propagates at normal incidence through a multi-layered wall is adapted to study transmission of speech signals. The algorithm, a time domain numerical implementation of the method of characteristics, is used to produce a digitized waveform of the exiting signal corresponding to tan incident signal consisting of a speech fragment. To assess the effects of distortion resulting from internal reflections, a speech pattern is digitized with the aid of standard multimedia tools and used as the input to the transmission algorithm. The transmitted signal is computed and converted to a convenient audio format for playback. To ensure that distortion in the output signal is truly effect of the medium, the signal is processed again using a noise reduction routine.
The developed theory, the modifications required for sampled speech
data, the simulations required to study the effects of the wall properties,
the applied signal processing techniques, and conclusions on the methodology
developed and the complexity in correlating objective wall properties and
subjective nature of speech intelligibility will be discussed.