Voice Recognition Using K-Means Clustering Based on Hidden Markov Model
Aloysius Bagas Pradipta Irianto · 2019
Ortolan Bunting bird has various song-types. For time being, there are not so many systems which automatically cluster the song-type of Ortolan Bunting bird. Speech recognition is one example of technology development in the 20thCentury which uses voice as the input. This research builds an automatic system to cluster the song-type of Ortolan Bunting bird automatically using K-means clustering algorithm based on Hidden Markov Models (HMM). Commonly, HMM is used on signal processing and for speech and speaker recognition on human voice. This research uses 5 song-types of bird, and every song-type is represented by 100 songs. The data will be the clustering using K-means based HMM with 3 feature extractions and the result validation employs index stability computation. The clustering parameter groups the data into 5 clusters with dissimilarity value of 8.2 %. It means that for several times of experiments there are 8.2 % data clustered into different group and 91.8% of cluster members which always gather in the same cluster. The highest similarity level was obtained using Greenwood function cepstral coefficients delta-acceleration (GFCC_D_A) feature with 36 parameters.