[29425] Artykuł: Preliminary processing of the human voice recordingsCzasopismo: Conference Archives PTETiS Tom: 31, Strony: 275-279Opublikowano: 2012 Autorzy / Redaktorzy / Twórcy
Grupa MNiSW: Pozostałe publikacje (niepunktowane) Punkty MNiSW: 0  Pełny tekst |
Analysis of the human voice is a difficult process. This is due to the complexity of the voice signal, and its individuality against every human being. Keep in mind that a person can freely change the tone of voice, speed of speaking and many other parameters that are independent of content of speech. The first task of the researcher is to eliminate differences in the voice signals for all recorded people. This is the only way you can get a sample of data that can be compared within a pool of recordings performed in the same way. This paper describes a proposal for the selection of audio data for a short speech on the basis of one voiced sound. The study was conducted a directional microphone and recorder having the ability to save as uncompressed audio. Recordings were carried out 31 people, and then the Fourier analysis, spectrogram and cepstrogram of selected data were determined. These characteristics were used to develop a path for dealing with voice recordings of differing length, amplitude level and a way of expression of the recorded persons. The study was based on comparison of the graphical results for various parameters of selection. The result is a strictly specified path of conduct, which can be used for the short-term analysis for most sound recordings conducted under different environmental parameters.
