Абстрактный

Accent Recognition using MFCC and LPC with Acoustic Features

Reena H. Chaudhari, Kavita Waghmare, Bharti W. Gawali

India is a multilingual country where 29 individual languages as having more than 1 million native speakers. Hindi & English are the official languages of the Republic of India. Hindi is the most widely spoken language in India. Every language has its own sound structure, grammar syntax and intonation pattern, which makes it unique. Speech is a vocalized form of human language and it is a primary means of communication between people. The accent is a significant component of any speech. An accent is the way that particular person or group of people pronounce words or sounds. It typically differs in the tone of voice, pronunciation and distinction of vowels, consonants, stress and prosody. The purpose of this research is to determine the impact of Hindi language on other Indian languages like Marathi, Marwadi and Urdu Language with accent and some other features. It uses Mel Frequency Cepstral Coefficients (MFCC), Linear Productive Coding (LPC) to extract speech features from four different language groups, fundamental frequency Formant (F0) and energy feature vectors are used to examine the difference between language groups. This experiment tested on a database having 10 Hindi sentences which are let out by native speakers of Hindi, Marwadi, Marathi and Urdu. The observation from experiment indicates that both techniques give accurate and identical outcome. F0 and Energy parameter are founded effective in Urdu Dataset.