Bird Species Classification Based on Mixed-GMM
Abstract
A bird species classification method is raised on the basis of Mel-Frequency Cepstral Coefficients (MFCC) and mixed Gaussian Mixture Model (mixed-GMM). The mixed-GMM is established according to the species of birds by classifying each bird recordings into call-song model (CSM) and male-female model (MFM), and extracting each species’ feature parameter respectively. In the recognition phase, each unknown bird recording’ MFCC is matching with the known CSM and MFM, and to classify it into the species which gets the highest score with its model. Results show that the accuracy rate of mixed-GMM is higher than conventional GMM.
Keywords
Mel-Frequency Cepstral Coefficients (MFCC), Mixed Gaussian Mixture Model (mixed-GMM), Call-song model (CSM), Male-female model (MFM)
DOI
10.12783/dtcse/wcne2016/5144
10.12783/dtcse/wcne2016/5144
Refbacks
- There are currently no refbacks.