Difference (Distance) Analysis

In dBSONIC a distance spectrogram can be computed from the time-frequency analyses of two sounds (reference and test analysis). The distance spectrogram shows differences between the two sounds in the time-frequency domain. From the distance spectrogram a mean distance spectrum and the distance versus time can be computed. Also the metrics mean distance and maximum distance are computed.
Distance computation can be based upon all dBSONIC time-frequency analyses (except Auditory Contours):

• Fourier (FFT) Spectrogram
• Auditory Spectrogram
• Third-Octave Spectrogram
• Wavelet Scalogram
• Modulation Spectrogram
• Prominence Spectrogram
• Specific Loudness
• Specific Loudness (Cores)
• Specific Fluctuation
• Specific Roughnesss

The distance analysis offers a gain and delay compensation. When comparing two versions of a sound the sounds are often delayed to each other. This may be due to the recording conditions or one sound may be processed by a system that introduced a certain overall delay. This delay in most cases is perceptual irrelevant but leads to very high differences when different time sections of the two sounds are compared.

In the following figure the Auditory Spectrograms of a door slam noise and of a processed (delayed) version of that sound are compared.