consonants include the six oral stop consonants /p, b, t, d, k, g/ and a glottal stop /p/. Thus, the following section (§5.2) describes the phonetic properties of Assam Sora nasal consonants with the help of acoustic analysis, then §5.3 describes the six oral stops with particular reference to the characteristics of voicing contrast among the six oral stops in Assam Sora. Subsequently,§5.4 describes the phonetic realization of the glottal stop in Assam Sora, and §5.5 concludes the chapter by summarizing the major findings of this chapter.
consistently at around 200 Hz for all nasals, other formants also occur for the nasals (Stevens, 2000). Moreover, apart from the formant frequencies, anti-formants are also observed in nasals. Prominent anti-formants arise in nasals mainly due to an oral coupling with nasal sounds. For example, in case of a bilabial nasal, the closed oral cavity acts as a side-branch and the side branch takes out the energy of the main tube. This results in creating of anti-formants in the bilabial nasal. Significantly, due to these anti-formants amplitude of the nasal formant frequency is lowered in nasal sounds. Additionally, anti-formant characteristics also vary with place of articulation of nasal sounds. For instance, anti-formant characteristics of alveolar and velar nasals are much lower than the bilabial nasal, since in alveolar and velar nasals the side oral cavity is much smaller then the side oral cavity of the bilabial nasal. Thus, prominence of nasal formants and anti-formants are generally taken into consideration for the classification of nasal consonants. Therefore, the classification of the nasals in Assam Sora is done by examining the nasal formants and the anti-formants.
In order to examine the acoustic characteristic of nasal sounds in Assam Sora, a data set consisting words with nasal codas including /m/, /n/ and /N/ is considered.
For /m/, 320 tokens; for /n/, 79 tokens and for /N/, 638 are extracted from the speech database collected from all four regions of Assam Sora. From the nasal data, spectrum is extracted for each nasal at the middle of the total duration of the nasal sounds in the coda position only using a Praat script. In order to do this, each sound file containing nasal sounds is transformed to a spectrogram using Praat (Boersma &
Weenink, 1992) with a Gaussian window of 5 ms window, 2 ms time steps, and 20 Hz of frequency step with a maximum frequency of 5000 Hz. Following this, spectrum at the mid point of the total duration of every nasal is calculated and power of the spectrum is derived using Equation 5.1, where r and i are the real and imaginary components, respectively, of the Discrete Fourier Transform of the speech signal.
Moreover, power of the spectrum is calculated in the 0 – 5000 Hz frequency range at
115 equidistant points, resulting in an interval of 43.5 Hz. Subsequently, the averaged spectrum across all nasal tokens is computed and displayed as seen in Figure 5.1.
pwr= 10×(log10(r2+i2)) (5.1) Figure 5.1: Average nasal spectra of Assam Sora with formants indicated with up arrows and anti-formants indicated with down arrows
Figure 5.1 shows the occurrence of a prominent nasal formant at around 250 Hz, as indicated by an up arrow. Similarly, another prominent formant is noticed at around 2500 Hz. It is noteworthy that there is a considerable difference between the amplitude of the first formant and the second formant. This difference is caused by the presence of anti-formants just preceding the second formant as indicated with a down arrow. Similarly, another anti-formant realized at about 3700 Hz has dampened the following spectrum almost flat. While the spectral peaks are clearly visible in the averaged spectrum of /m/, /n/ andN, the side-branch induced differences among the three nasals are not noticeable. The reason for this maybe due to the averaging of the nasal spectrum that has smoothed off any place of articulation specific (or side- branch specific) differences. Therefore, in order to see the characteristics of the nasal spectrum, without spectral characteristics being smoothed out, single spectra of a female speaker producing three nasal sounds /m/, /n/ and /N/ is plotted in Figure
5.2.
Figure 5.2: Nasal spectra of Assam Sora as produced by a single speaker
-120 -100 -80 -60 -40 -20 0
0 131 261 392 522 653 783 914 1044 1175 1305 1436 1566 1697 1827 1958 2088 2219 2349 2480 2610 2741 2871 3002 3132 3263 3393 3524 3654 3785 3915 4046 4176 4307 4437 4568 4698 4829 4959
dB
Frequency in Hz
m n ŋ
From Figure 5.2 it is evident that, all three nasals show a formant peak at about 200 Hz. The bilabial consonant /m/ shows prominent spectral peaks at 260 Hz, 2500 Hz and 4300 Hz. Another peak at about 1500 Hz is dampened by anti-formants approximately at 1300 Hz and 2000 Hz. In case of the alveolar nasal /n/, only the first nasal peak is the prominent one whereas, potential nasal peak at about 2500 Hz is dampened by anti-formants. The spectrum for the velar nasal /N/ shows a clear nasal formant at around 250 Hz and its harmonics are also quite prominent. However, no other prominent nasal peaks are noticed in the spectrum apart from the harmonic peaks. In case of the bilabial nasal /m/, there are two noticeable anti-formants at about 2000 Hz and at about 3500 Hz. Similarly, in case of the alveolar nasal /n/
a prominent anti-formant is noticed at about 4000 Hz. In case of velar nasal /N/
also, an anti-formant around 4000 Hz is prominent. Hence, considering, Figure 5.1 and Figure 5.2, it is can concluded that in case of Assam Sora nasals, nasal peaks are predominantly notice between 200−250 Hz and another nasal peak is noticed between 2500−3000 Hz. Similarly, a prominent nasal anti-formant is also noticed between 3500−4000 Hz. hence, from these nasal peaks and prominant anti-formants,
nasal consonants in the word final position can be identified in a continuous speech in Assam Sora.