A new similarity scale called the Geometric Distance, that numerically evaluates the degree of likeness between two patterns including character and voiceprint patterns, is proposed.

Human beings, dogs, cats, and other such animals have “the sense of similarity” in hearing and sight.
To realize “the sense of similarity” using an algorithm called “similarity scale” is an important subject for developing computer intelligence.

The similarity is often measured as a “distance” between the two patterns.
The similarity scale works as follows: for character and voiceprint patterns for which a researcher would recognize two patterns as similar to each other, the computer software outputs a small value, and for character and voiceprint patterns for which a researcher would recognize the two patterns as dissimilar, the computer software then outputs a large value.

In sound/image recognition, a known spectrum(or image) stored in PC memory is called here the “standard pattern”, and a comparison spectrum(or image) is called the “input pattern”.
The degree of likeness between the standard pattern and the input pattern is evaluated using a similarity scale.
If the similarity of the standard and input patterns is close, then those two patterns are considered to be in the same category and the input pattern is recognized.

Application Field:
Abnormal sound detection, Voice recognition, Image recognition, Convolution for deep learning, Recognition of wild life calls.

Key Words:
Similarity measure, Similarity function, Similarity scale.

The GD Similarity is our product name of geometric distance.