Method and apparatus for analyzing animal vocalizations, extracting identification characteristics, and using databases of these characteristics for identifying the species of vocalizing animals

A method for identifying the species, appearance or other characteristics of an animal based on vocalizations, comprising:

• a. capturing the sound of an animal vocalization onto a digital device as sampled digital audio data;
• b. performing signal analysis on the sampled digital audio data to extract a plurality of analysis features comprising one or more of:
• spectral peak, spectral centroid, spectral variety, spectral harmonicity, RMS amplitude, RMS delta, and spectral peak delta, which describe low-level time-domain and/or frequency-domain properties of the sampled digital audio data;
• c. using the plurality of analysis features to perform audio onset detection, event delineation and event grouping to determine break-points in time that describe individual time-segments or notes (hereinafter “Elements”) and groups of Elements, (hereinafter “Phrases”) of the vocalization captured in the sampled digital audio data;
• d. using the plurality of analysis features and audio onset detection/segmentation results, perform Element and Phrase analysis to extract a plurality of high-level symbolic features, the plurality of defined high level Element/Phrase features comprising one or more of:
• symbolic duration and/or inter-onset (IOI) ratio,
• symbolic loudness,
• symbolic pitch and pitch span,
• symbolic tone-color quality chosen from a plurality of defined qualities comprising one or more of noise, clear, formant, complex, and buzz,
• symbolic pitch shape and profile chosen from a plurality of defined profiles and shapes comprising one or more of steady, slur-up, slur-down, bow-up, bow-down, trill, vibrato and chaotic,
• symbolic amplitude envelope shape and profile chosen from a plurality of defined profiles and shapes, comprising one or more of steady, percussive, ADSR, swell, tremolo and chaotic, and
• symbolic time and pitch intervals between the Elements and/or Phrases;
• e. using information in the plurality of low-level analysis features and high-level symbolic features to identify a match within a collection of similar features extracted or derived from vocalizations of a plurality of animals of a plurality of known species.