PERFORMANCE ANALYSIS OF SPEECH WAVEFORM CODECS USING A DUAL SINUSOID INPUT
Date
1983-05-02
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
ORCID
Type
Degree Level
Masters
Abstract
The telephone industry in North America currently uses a 1-kHz sinusoid input to analyze the performance of speech encoder-decoders (codecs). This thesis shows that a signal composed of the sum of two sinusoids of equal amplitudes at 1 kHz and 1.1 kHz provides a better indicator of waveform codec performance with speech than the single sinusoid. The dual-sinusoid signal possesses more speech—like amplitude and slope distributions than the single sinusoid.
SNR's were measured with single-sinusoid, dual-sinusoid and speech inputs to a collection of several waveform codecs comprising /1-law 15-segment Companded PCM, ADM, and several ADPCM codecs with adaptive quantizers. Two SNR measuring techniques were used, one which obtained the quantizing noise signal by notching the test signal component from the codec output and the second which obtained the noise signal by subtracting the codec input and output. The latter waveform subtraction method required the design and construction of a laboratory test instrument. The SNR results indicated that when compared to the single
sinusoid, the dual sinusoid measurements were more consistent with those for speech.
Description
Keywords
Citation
Degree
Master of Science (M.Sc.)
Department
Electrical Engineering