Politehnika i dizajn, Vol. 7 No. 1, 2019.
Izvorni znanstveni članak
https://doi.org/10.19279/TVZ.PD.2019-7-1-06
NEGATIVE CLASSIFICATION DIGITAL INFORMATION OF VOICE TONE
Tomislav Jarimić
orcid.org/0000-0003-0470-3523
; Fakultet organizacije i informatike Varaždin, Sveučilište u Zagrebu, Varaždin
Sažetak
In a technical sense, does not exist a developed
recognition of voice tone on a human level. Voice
recognition and speech recognition are concepts
that are completely different in content. The
purpose of speech recognition is to recognize
a speaker's speech, while recognizing a voice
identifies the known person or an unknown
person. The voice tone of a person is a key pattern
in the voice of the speaker. It can be seen as a
subjective sense of the volume, height and color
of the sound and its perception in the space. The
aim of the research is to find in the recorded
voice of the speaker a characteristic feature that
can determine the digital information of the of
the voice tone. By using digital information of
voice tone and using information-communication
technology, we recognize a gender, a target
speaker, or a group of speakers. Intention is to
use Walsh-Hadamard's coefficients to determine
the characteristic features. In the preprocessing
process, we separate the characteristic features
for digital information of the voice tone. With
method of the Cepstral analysis determines the
basic frequency F0
for each recorded sound
on which basisi the respective harmonics are
calculated. The obtained samples are classified
by gender. In the next step from the recorded
sound, the Walsh-Hadamard coefficients are mark
off, classified by gender, and the results of the
classification are compared. In practice, the model
is used to identify, or to quickly eliminate a target
group of data.
Ključne riječi
digital information of voice tone; improvement of voice tone characteristic; Walshhadamard coefficients
Hrčak ID:
220657
URI
Datum izdavanja:
30.4.2019.
Posjeta: 888 *