K-mean Clustering and Arabic Vowels Formants Based Speaker Identification System (original) (raw)

This paper introduces and addresses the proposes of a new approach for speaker feature extraction based on experimental and theoretical approached, where the formants of Arabic Vowels are proposed to distinguish the speaker features from each other. Discrete Wavelet Transform (DWT) in conjunction with Algorithmic Power Spectrum Density (PSD) is used to illustrate the distinguisher of different speaker formants. This approach provides a more efficient method in speaker recognition rate, i.e., higher accuracy. Kmeans clustering (KC) and Root Mean Square Difference Similarity Measure (RDSM) are used for features classification. Instead the conventional method extracts the features from one word or more. In this paper the authors proposed a new method to utilize the Arabic Vowels. Ultimately, the attained results by the presented method showed considered a performance in classification, which reaches about 94% in classification rate. As a result of DWT utilization, the system works with...