Browse Prior Art Database

# Use of Symmetrized Dot-Patterns to Produce Polar Graphs of Speech Waveforms

IP.com Disclosure Number: IPCOM000044456D
Original Publication Date: 1984-Dec-01
Included in the Prior Art Database: 2005-Feb-06
Document File: 2 page(s) / 85K

IBM

## Related People

Pickover, CA: AUTHOR

## Abstract

The described symmetrical polar graph dot-patterns provide a novel visual characterization useful in speech analysis, voice identification, and waveform characterization in general. Let the amplitude of a waveform at a time point j be symbolized as Yj . The algorithm maps the amplitude at points j and j+l into r and 0 in polar space, respectively, in the following manner: rj = Yj j = 1,2,3,.,N - 1 0 = i + Yj+1 i = 0,60,120,180,270,360;j = 1,2,3,.,N - -1ij 1 0 = i-Yj+1 i = 0,60,120,180,270,360;j = 1,2,3,.,N - 1 -2ij where N is the number of points in the pcm waveform. For each iteration of the program, an (r, 0-1) and (r, 0-2) is displayed on the video screen. In other words, the resultant figures have 6 mirror planes (at multiples of 60 degrees), and 1 input amplitude point would give rise to 12 display points.

This text was extracted from a PDF file.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 52% of the total text.

Page 1 of 2

Use of Symmetrized Dot-Patterns to Produce Polar Graphs of Speech Waveforms

The described symmetrical polar graph dot-patterns provide a novel visual characterization useful in speech analysis, voice identification, and waveform characterization in general. Let the amplitude of a waveform at a time point j be symbolized as Yj . The algorithm maps the amplitude at points j and j+l into r and 0 in polar space, respectively, in the following manner: rj = Yj j = 1,2,3,.,N - 1 0 = i + Yj+1 i = 0,60,120,180,270,360;j = 1,2,3,.,N - -1ij 1 0 = i-Yj+1 i = 0,60,120,180,270,360;j = 1,2,3,.,N - 1 -2ij where N is the number of points in the pcm waveform. For each iteration of the program, an (r, 0-1) and (r, 0-2) is displayed on the video screen. In other words, the resultant figures have 6 mirror planes (at multiples of 60 degrees), and 1 input amplitude point would give rise to 12 display points. This high degree of symmetry is useful in aiding the human analyst in detecting patterns which characterize specific waveforms, as described below. The gain is normalized for each pattern, so overall amplitude is not a factor in the characterization. The Figs. 1A - 1D include examples of the use of this method on synthetic and human speech sounds. For the plots, the input waveform is 50 ms in duration, and the sampling rate is 10 KHz. Each sound bears a distinct signature which is independent of input data window- length or position in time in the utterance. These patterns characterize both the speaker and the phoneme being spoken. In addition, patterns are not obscured by the overall pitch of the input sound. In summary, the representation provides the r...