PERBANDINGAN AKURASI ANTARA MEL-FREQUENCY CEPSTRAL COEFFICIENT (MFCC) DAN LINEAR PREDICTIVE CODING PADA APLIKASI PENGENALAN SUARA
Total View This Week0
Total View This Week0
Institusion
Institut Teknologi Perusahaan Listrik Negara
Author
HANIF, M. NURY
Haris, Abdul
Putra, Eka
Haris, Abdul
Putra, Eka
Subject
Teknik Informatika
Datestamp
2022-09-21 08:21:44
Abstract :
This study discusses the application of voice recognition voice extraction methods for comparing between Mel-Frequency Cepstral's (MFCC) and Linear Predictive Coding (LPC) to determine the most accurate method. As for the problems that emerged at the time of voice recognition is the lack of inaccuracies in the introduction, so need to be examined again what method has high accuracy and computational time short. The comparison is done in a way to compare the stages of the process of each of these methods. In this thesis, the author compares the four stages of each method starting from the stage of pre-emphasis to clarify the vocal sound and eliminate defined high, continued with the framing to split into several blocks or frames, and continued with the process of windowing to eliminate noise and smoothing each block that is already divided and the last process is the FFT and autocorrelation. Other features in the making this comparison application will bring up the final results of the perandingan data accuracy tested with a trainer as well as data showing long computation of both these methods in order to find out the fastest method and effective for voice recognition. The making of the application method of comparison and MFCC LPC was created using matlab R2017A tools application then tested using the method of the User Acceptance Test, with test divided into four (4) main characteristics Data End User communications, Efficiency, Operational Ease, Reusability and reliability analysis of results obtained 87%. This system is classified into the category of good. And the end results of this comparison test is a method of extracting sound level MFCC accuracy and computational time faster than the extraction of voice LPC
This study discusses the application of voice recognition voice extraction methods for comparing between Mel-Frequency Cepstral's (MFCC) and Linear Predictive Coding (LPC) to determine the most accurate method. As for the problems that emerged at the time of voice recognition is the lack of inaccuracies in the introduction, so need to be examined again what method has high accuracy and computational time short. The comparison is done in a way to compare the stages of the process of each of these methods. In this thesis, the author compares the four stages of each method starting from the stage of pre-emphasis to clarify the vocal sound and eliminate defined high, continued with the framing to split into several blocks or frames, and continued with the process of windowing to eliminate noise and smoothing each block that is already divided and the last process is the FFT and autocorrelation. Other features in the making this comparison application will bring up the final results of the perandingan data accuracy tested with a trainer as well as data showing long computation of both these methods in order to find out the fastest method and effective for voice recognition. The making of the application method of comparison and MFCC LPC was created using matlab R2017A tools application then tested using the method of the User Acceptance Test, with test divided into four (4) main characteristics Data End User communications, Efficiency, Operational Ease, Reusability and reliability analysis of results obtained 87%. This system is classified into the category of good. And the end results of this comparison test is a method of extracting sound level MFCC accuracy and computational time faster than the extraction of voice LPC
Download
book
BibTex
Latex, Jabref
cloud_download
Original Resource
url resource Institution
