dc.description.abstract | Researches on voice signals have been carried out using various signal processing methods, one of them known as Mel Frequency Cepstrum Coefficients (MFCC) which is based on Fourier transformation. MFCC as feature extraction combined with Artifical Neural Network (ANN) as classifier has been shown to have a very good accuracy. Based on previous research, this research is conducted with more complicated cases with the various medical conditions of the speaker. Bad medical conditions, such as cold, is a factor that can affect the human voice, making it more complicated to recognize using signal processing methods. From this research, it can be concluded that the use of MFCC method as the feature extraction combined with ANN as the classifier proved to have a good capability to recognize words from speakers with bad medical conditions. The accuracy that has been obtained is 99% using backpropagation neural network with 50 hidden neurons and using the Traingdx training methods. | en |