african elephants

Patrick Clemins

Project Goal

A Hidden Markov Model (HMM) system is used to automatically classify African elephant vocalizations. The development of the system is motivated by successful models from human speech analysis and recognition. Classification features include frequency-shifted Mel-Frequency Cepstral Coefficients (MFCCs) and log energy, spectrally-motivated features which are commonly used in human speech processing. Experiments, including vocalization type classification and speaker identification are performed on vocalizations collected from captive elephants in a naturalistic environment. The system classified vocalizations with accuracies of 94.3% and 82.5% for type classification and speaker identification classification experiments, respectively. Classification accuracy, statistical significance tests on the model parameters, and qualitative analysis support the effectiveness and robustness of this approach for vocalization analysis in non-human species.


Noisy Rumble
Rev (usually receeds a rumble)

Classification Results


79.7% Accuracy

Call-Type on Clean Data

91.4% Accuracy

Speaker Identification

84.6% Accuracy

Estrous Cycle

74.5% Accuracy

Behavior Correlation on Rumbles

51.3% Accuracy (Not Statistically Signicifant)

MANOVA Results

Call Type F152,5051 = 73.337, P<0.001
Speaker Identification F190,13426 = 52.089, P<0.001
Estrous Cycle F76,9172 = 13.190, P<0.001
Behavior Correlation F38,3723 = 7.329, P<0.001