By its very nature, the spectrum of a person's voice is quite complex. Because the spectrum changes rapidly with time, we will use what is called "time-frequency" analysis techniques to characterize the voice.
Views: Spectrogram, Spectrum
Sampling rate: 22050 Hz
FFT size: 512 or 1024
FFT overlap percentage: As detailed below.
Smoothing window: Hamming
Averaging size: 2 or 3
Amplitude Scaling: Logarithmic
Frequency Scaling: Linear
Sampling Fromat: 16 bit mono.
Mic Compensation: As required.
Spectral weighting: Flat
Record a short time segment of the voice to be analyzed.
Change to the Post-Processing mode.
Initially set the Overlap Percentage to zero.
Click the rewind button to position the current file position to the start of the file.
Run the analyzer until the spectrogram display is filled or the end of file position is reached.
Identify the voice segment to be analyzed in detail by highlighting a portion of the spectrogram and play it back through the speakers.
Use the left mouse button to readout the time corresponding to the start of the voice segment under analysis.
Use the scroll bar on the main application toolbar to adjust the current file position to the start of the desired time segment.
Set the overlap percentage to 90%
Run the analyzer and examine the spectrogram view.
Readjust the overlap percentage to expand/contract the total time displayed on the spectrogram time axis to give you the desired detail.
Double click at any point on the spectrogram to place the spectrum corresponding to the cursor position in the spectrum view. (if the time series view is displayed, the corresponding time series data will be displayed as well)
You can open the spectrogram display options dialog box and adjust the plot top, plot range and frequency span to optimize the display. The spectrogram will be recomputed with the new settings when the dialog box is closed.
The same procedure applies when using the 3-D surface plot in place of the spectrogram.