Documente Academic
Documente Profesional
Documente Cultură
REZUMAT ABSTRACT
Societatea actual este o societate bazat pe realizri The society we live in is a society based on high quality
tehnologice informaionale de nalt calitate, tot mai information technology accomplishments, more and more
multe fiind domeniile n care sunt utilizate sisteme
are the fields where high precision computerized tools
computerizate de mare precizie. Unul dintre acestea, care
se afl n plin ascensiune, l constituie studiul vocii are employed. One of these fields, which is constantly
cntate. growing, is the study of the singing voice.
Evaluarea parametrilor vocali poate fi efectuat, aadar, Evaluation of vocal parameters may be done by specially
prin intermediul unor programe special create. Lucrarea created software. This study presents an attempt to
de fa prezint o ncercare de a sintetiza i de a trece synthesize and analytically filter the vocal sound by
printr-un filtru analitic computerizat sunetul vocal, means of a computer using Praat software.
folosind n acest scop programul Praat.
Key words
Cuvinte cheie
Singing voice, Praat program, fundamental frequency,
Voce cntat, Programul Praat, frecven fundamental, spectrogram, pitch, formant.
spectograma, nlimea sunetului, formant.
FOREWORD
INTRODUCERE
Among the study fields based on web technologies
Printre domeniile de cercetare bazate pe utilizarea
tehnologiilor web legate de studiul vocii se numr related to the study of the voice we can count the
analiza i evaluarea acustico-fiziologic a mecanismului analysis and acoustic-physiologic evaluation of the
vocal. Obiectivele unor asemenea cercetri sunt vocal mechanism. The objectives of this kind of
monitorizarea calitii sunetului emis, precum i gsirea research are to monitor the quality of the vocal sound,
unor soluii de natur fiziologic n vederea asigurrii and also to find some solutions of physiologic nature
longevitii aparatului vocal i reducerii riscurilor de in order to ensure longevity of the vocal chords and to
disfuncionalitate. reduce the risks of malfunction.
Unul dintre avantajele sistemului computerizat de
One of the advantages of the computerized system of
investigaie l reprezint caracterul su obiectiv,
reproductibil, precum i accesibilitatea i uurina cu care investigation is represented by its objective and
poate fi folosit. Bazele de date create prin intermediul reproducible character, as well as its accessibility and
acestor programe ofer astfel un reper standardizat ce ease of use. Databases created by these programs offer
poate deveni un punct de referin pentru cntreii a standardized landmark which can become a reference
nceptori i un sprijin pentru profesorii de canto n point for beginners and a tool for voice teachers.
domeniul cultivrii i dezvoltrii vocii cntate.
THE PRAAT PROGRAM AN
PROGRAMUL PRAAT O UNEALT ADVANCED TECHNOLOGICAL TOOL
TEHNOLOGIC AVANSAT
Ultimele decenii se remarc prin apariia mai multor The last decades are remarkable because of new
programe destinate studiului vocii (vorbite i cntate). programs designed to study the human voice (spoken
Printre acestea amintim programul PitchXtractor pentru and sung). Among these we find PitchXtractor for
Macintosh, care a fost creat anterior lui Praat, i este un Macintosh, which was created before Praat and is less
precursor mai puin performant n comparaie cu acesta. successful than it.
Un alt program, de mare complexitate din punct de
Another program, very complex from the point of view
vedere al posibilitilor de analiz a semnalului vocal, l
reprezint sistemul VOCSPEC, creat n cadrul clinicii of its analytical possibilities, is VOCSPEC, created by
ORL Cluj, n colaborare cu SOFTWARE ITC Cluj. Din
pcate, din motive tehnice, acest program nu mai poate fi
65
Tehnologii informace i de comunicaie n domeniul muzical
[ [
folosit. Aadar, Praat rmne la ora actual cel mai the Otolaryngology Clinic from Cluj-Napoca together
performant program de analiz a parametrilor vocali, with SOFTWARE ITC Cluj. Unfortunately, for
care poate fi accesat gratuit la adresa: technical reasons, this program cannot be used
http://www.fon.hum.uva.nl/praat.html
anymore. So Praat is now the best program for
AVANTAJE I POSIBILITI analyzing vocal parameters, and can be downloaded
Programul Praat este un program multifuncional, freely at: http://www.fon.hum.uva.nl/praat.html.
punndu-ne la dispoziie o palet diversificat a
instrumentelor informaionale de investigaie, care s ADVANTAGES AND POSSIBILITIES
permit analiza n detaliu a segmentelor muzicale The Praat program is a multifunctional program,
examinate. Printre opiunile disponibile se numr which allows us to investigate on multiple accounts
editarea de sunet, redarea, reprezentarea grafic,
the vocal sound and to analyze recorded pitches in
modificarea, sintetizarea, conversia, filtrarea, precum i
generarea de sunet. Cercetarea poate fi fcut pe dou detail. The options available are editing sound,
nivele, folosind: playing, graphic representation, modifying,
x tehnici de cercetare de baz sau synthesizing, conversion, filtration, and also sound
generating. Research can be done on two levels:
x tehnici avansate de cercetare
x basic technical research or
Fiierele care pot fi introduse n program trebuie s fie
nregistrate n sistemul mono, nregistrrile stereo fiind x advanced technical research
mprite n mod automat pe canale. Fiierul va conine o The files that can be opened need to be mono files,
singur voce. Muzica pe mai multe voci sau nsoit de
un acompaniament instrumental consistent nu va putea fi because stereo recordings will be automatically split
analizat dect n condiiile separrii acestora. Ulterior, into channels. The file will contain only one voice.
vocile pot fi reasamblate ntr-o singur imagine cu Multiple voice or instrumental accompaniment could
uurin. not be analyzed without removing them. Subsequently
Important de reinut este faptul c sistemul Praat a fost they all can be merged together in one file.
conceput ca un sistem de analiz asupra vocii vorbite i It is important to know that Praat was devised for
cntate i mai puin ca un mijloc de investigaie asupra
liniilor instrumentale. De aceea, cea mai potrivit analyzing spoken and sung voice, and not for
utilizare a sa o reprezint studiul asupra vocii umane. investigating instrumental lines. Therefore its best use
is with the human voice.
Figura 1. Editarea sunetului prin intermediul Figure 1. Editing sound with the Praat program
programului Praat
The upper half shows the graphic contour of the recorded
melody (the sound's waveform), and the lower part has
Jumtatea de sus a ecranului arat conturul grafic al
some written instructions. The vertical red line in the
melodiei nregistrate (unda sinusoidal a sunetului),
middle of the page can be relocated anywhere on the
partea de jos avnd instruciuni scrise. Linia vertical
graphic to play the recording. The program allows
roie, aflat n mijloc, poate fi relocat n oricare parte a
graficului, pentru a reda melodia. Programul permite selecting a portion of the vocal line for further analysis.
selectarea unei poriuni a liniei vocale pe care dorim s o We have selected a sample that contains the F note on the
analizm n amnunt. Selectm poriunea ce conine third octave (F6=1397Hz), which is sung on the vowel A.
sunetul Fa din octava 3-a (F6 = 1397Hz), intonat pe In order to show all the tools in one screen we choose the
vocala A. following:
n continuare alegem urmtoarele setri:
Figura 4. Spectograma
Analiza spectral de fa pune n eviden prezena Figure 4. Spectrogram
formanilor caracteristici intonrii vocalei A. Frecvena This spectral analysis shows the presence of the formants
densitii acustice a sunetului se concentreaz n jurul cifrei which are present when singing the vowel A. The
de 4862 Hz. n partea de sus a ecranului, acolo unde aceast
frequency of the acoustic density is mainly present
cifr este depit, observm micorarea densitii energiei
around 4862 Hz. On the upper side of the spectrogram we
acustice. n consecin, spectograma realizat ilustreaz
existena unor structuri armonice sau formani de mare observe a reduced energy of acoustic density. Thus, the
intensitate, dezvluind calitile bogate ale sunetului vocal spectrogram unveils high intensity harmonic structures or
emis n registrul acut al vocii de sopran. Urmtorul pas n formants, disclosing the rich harmonic qualities of the
completarea tabloului analitic conturat l constituie high pitch soprano voice. Next step is examining the
examinarea nlimii: pitch:
Conturul melodic este mai bine spaiat acum, fiind Pitch contour, now better spaced up, can be seen more
evideniat mai clar pe axa timp/nlime. Cadrul de clearly on the time/pitch axis. The inspection framework
examinare este cuprins n limitele valorilor de 75-
is between 75-1400 Hz. So, moving the cursor on any of
1400Hz. Astfel, oprind cursorul pe oricare dintre sunetele
dorite putem vedea nlimea exact a acestuia exprimat the sounds we can see its precise pitch in Hz as well as its
n Hz precum i momentul lui de desfurare precis temporal place (including milliseconds). Doing so we can
(pn la nivelul de milisecunde). Efectuarea acestei see that one of the highest pitches has a value of 1234,96
operaiuni indic faptul c unul dintre sunetele cele mai Hz and is produced at second 9,91071.
nalte, avnd valoarea de 1234,96Hz, este produs n
secunda 9,910971. Setting the lower threshold for the pitch screen is very
Setarea pragului inferior al nlimilor constituie o sarcin important. It represents a basic technical requirement
de mare importan. Aceasta reprezint o cerin tehnic because setting the threshold too low will lead to a loss of
de baz a programului, deoarece o setare prea joas ar fast changing F0, and setting it too high will lead to a loss
cauza pierderea schimbrilor foarte rapide ale F0, iar una
of lower F0 values. Knowing all this we have set the
prea nalt ar conduce spre pierderea valorilor foarte
joase ale F0. innd cont de toate acestea am setat pragul threshold at 75 Hz.
inferior al nlimii la valoarea de 75 Hz. A deeper analysis can be done using the advanced pitch
O analiz mai aprofundat poate fi realizat folosind settings. This option allows selecting a very short
setarea avansat a nlimii. Aceast opiune permite fragment, just a few seconds. We have selected the
selectarea unui segment foarte scurt, cu o durat de doar
section between 8,569112 and 12,120796 seconds and so
cteva secunde. Am selectat seciunea cuprins ntre
secunda 8,569112 i 12,120796, astfel durata poriunii the duration of this part is of 3,551684 seconds. In order
analizate este calculat de program la 3,551684 secunde. to understand the resulting graphic table we need to
Pentru a putea descifra tabloul grafic rezultat este nevoie understand the pitch detection algorithm used for this
de nelegerea algoritmului de detectare a nlimii care
program (and also for similar programs), as well as the
st la baza programului de fa (i a programelor
similare), precum i de familiarizarea cu noiunea de autocorrelation notion.
autocorelare. Spectral/temporal pitch detection algorithms are based on
Algoritmii de detectare a nlimii spectral/temporale sunt a computerized processing combination of the temporal
bazai pe o combinaie de procesare computerizat a domain using an autocorrelation function such as
domeniului temporal, folosind o funcie de autocorelare,
crosscorelation. So, to identify the pitch of a sound, the
precum corelarea ncruciat. Astfel, pentru a identifica
nlimea sunetului, este procesat domeniul frecvenei frequency domain is processed using spectral
fiind folosit informaia spectral. O pist final n information. A final track in calculating the pitch can be
calcularea nlimii poate fi bazat pe programarea based on dynamic programming through estimation of
dinamic prin intermediul estimrii candidailor din cele candidates from the two domains: temporal and
dou domenii: temporal i cel al frecvenei. Avantajul
frequency. The advantage of such an approach is based
unei asemenea abordri const n faptul c eroarea de
proiecie a unui domeniu va fi redus datorit procesrii on the fact that the projection error of a domain will be
celuilalt domeniu. reduced because of processing the other domain.
Algoritmul detectrii nlimii este un algoritm conceput Pitch detection algorithm is an algorithm designed to
pentru a estima nlimea sau frecvena fundamental a
estimate pitch or F0 of a quasi-periodic signal. Its value is
unui semnal cvasiperiodic. Valoarea acestuia este
invertit rezultnd calculul frecvenei. La baza inverted, resulting frequency calculation. The algorithm
algoritmului st o unealt matematic de depistare a is based on a mathematical tool of tracing repetitive
tiparelor repetitive (de exemplu a unui semnal periodic) patterns (for example a periodic signal) called
numit autocorelare. autocorrelation.
n procesarea semnalelor, autocorelarea reprezint un
For processing signals the autocorrelation represents a
domeniu ce poate oferi informaii despre evenimente
repetitive, precum sunt timpii muzicali sau frecvenele domain which can offer information about repetitive
pulsaiei. Autocorelarea poate fi folosit n estimarea events such as musical beats or pulsation frequencies.
nlimii unui ton muzical, sau la identificarea frecvenei Autocorrelation can be used in estimating the pitch of a
fundamentale lips n cadrul unui semnal dedus din musical tone, or to identify missing F0 from a signal
frecvenele sale armonice.
deducted from its harmonic frequencies.
Familiarizarea cu noiunile explicate mai sus permite
nelegerea urmtorului grafic: Being familiar with the notions explained above we will
understand the next graphic:
69
Tehnologii informace i de comunicaie n domeniul muzical
[
43,166236 secunde relev o diferen semnificativ n we can see a significant difference regarding the lower
ceea ce privete limita inferioar i mai puin pronunat threshold and a lesser difference regarding the superior
la nivelul pragului superior, ncadrndu-se ntre 50 i threshold, which is between 50 and 84,04 dB. The mean
84,04dB. Media segmentului de 3,55 sec. a fost calculat of the 3.55 second segment was calculated by the
de program la valoarea de 69,74dB. Aceast medie este program at 69,74 dB. We got this mean for a fragment
obinut ntr-un fragment cntat n registrul acut al vocii sung in the upper range of the soprano voice, where vocal
de sopran, unde penetrana vocal este datorat nu att penetration is due not so much to sound intensity as to
intensitii sunetului cntat ct, mai ales, capacitii
the singer's capacity to focus the sound. Considering
interpretului de focalizare a sunetului. Din acest punct de
these facts, special attention is paid to the analysis of the
vedere o importan deosebit revine analizei formanilor,
formants, which probably represent the best test for
care reprezint probabil cel mai amplu examen n
aprecierea calitilor acustice ale vocii. Tocmai existena appreciating vocal acoustic qualities. The existence of
formanilor confer vocii penetran i permite formants gives the voice its penetrating sound and allows
cntreilor s se aud peste acompaniamentul orchestral. the singers to be heard over an orchestra. The formants
Formanii reprezint armonicele sunetului amplificate de represent the sound harmonics amplified through the
cavitile de rezonan, care joac un rol definitoriu n resonant cavities, which play a crucial role in forming the
formarea timbrului vocal. vocal timbre.
Conform studiilor foniatrice (Murean, 2010) lungimea According to phoniatric research (Murean, 2010), the
tipic a tractului vocal este de 17-18 cm. Aceasta typical size of the vocal tract is 17-18 cm. It generates a
genereaz o frecven fundamental de 500Hz, F0 of 500 Hz, determining frequencies of the formants at
determinnd frecvene ale formanilor la 500, 1500 i 500, 1500 and 2500 Hz, which represents the limit of
2500Hz, care exprim limita frecvenelor observate. observed frequencies. The articulatory muscles which
Muchii articulatori acionai n producerea vocalelor function for vowel production are responsible for
determin modificrile de frecven ale formanilor. modifying the frequencies of formants. To raise the
Astfel, pentru a crete frecvena pe un sunet situat n frequency on a sound found in the upper vocal range the
registrul acut al vocii de sopran este nevoie de coborrea
lower jaw needs to be lowered, which is exactly what the
mandibulei, ceea ce a i fcut cntreaa n timpul
soprano did while recording this aria. Upon lowering the
nregistrrii ariei. Datorit coborrii mandibulei, primul
jaw, the first formant (F1) was substantially raised. The
formant (F1) a putut fi crescut n mod substanial. O
vowel is also very important. Mozart is again a genius
importan deosebit revine i vocalei cntate. Mozart se
dovedete din nou un mare cunosctor al calitilor vocii indicating the vowel A to be sung by the singer.
umane i indic vocalizarea lui A. The formant graphic generated by the Praat program
Graficul formanilor obinut prin intermediul programului shows a rich variety of formants, proving the outstanding
Praat ilustreaz existena unei bogate palete, dovedind qualities of the singer. This picture offers a model which
calitile vocale deosebite ale sopranei. Acest tablou every soprano would like to reproduce for a quality
ofer aadar un model spre care poate tinde un cntre performance:
dornic s se afirme pe scenele lumii:
71
Tehnologii informace i de comunicaie n domeniul muzical
[ [
analiz a sunetului. Asemenea demersuri pot avea un process as well. One of the research areas of the second
impact deosebit att asupra procesului de half of the 20th century undertook to explore the
instruire/formare ct i asupra domeniului componistic interiority of the sound, capitalizing on the natural
muzical. Una dintre orientrile muzicii din a doua resonance. The idea of editing the sound has proliferated
jumtate a secolului XX i-a propus explorarea among musicians, new research being done in exploring
interioritii sunetului prin revalorificarea rezonanei sale the natural harmonic spectrum. Scientific approach of the
naturale. Ideea prelucrrii sunetului a creat o emanaie sound algorithms can further research in this area.
puternic, compozitorii cutnd i descoperind modaliti
personale de explorare a spectrului armonicelor naturale.
Explorarea tiinific, algoritmic a sunetului poate crea
un nou impuls n direcia alimentrii unor asemenea
tendine componistice.
BIBLIOGRAFIE / REFERENCES
[1] Johnson, A. M., i Kempster, G. B., ( 2011). Classification of the classical male singing voice using
long-term average spectrum. Journal of Voice 25(5), 538-543.
[2] Murean, R., Chiril, M., (2010). Reabilitarea i igiena vocii. Cluj-Napoca: Alma Mater.
[3] Sataloff, R. T., (1991). Professional voice: the science and art of clinical care. New York: Raven.
[4] Lamarche, A., Ternstrm, S., i Pabon, P,. (2010). The singer's Voice Range Profile: Female
professional opera soloists. Journal of Voice, 24 (4), 410-426.
72