Applied Speech and Audio Processing: With matlab examples
Download 2.66 Mb. Pdf ko'rish
|
Applied Speech and Audio Processing With MATLAB Examples ( PDFDrive )
4.2. Psychoacoustics
67 the masking effect has frequency dependence based on the frequency of the masking tone [4,14]. 4.2.11 Frequency discrimination Frequency discrimination by humans is dependent upon absolute frequency, and to some extent amplitude, however an approximate figure would be 2 Hz for a 65 dB SPL signal at 1 kHz. Thus a tone of 1002 Hz can just be distinguished from a 1000 Hz tone by an average listener. Frequency discrimination is related to pitch perception which decreases with increas- ing amplitude for tones below about 2 kHz, but increases with increasing amplitude for tones above about 4 kHz. Unless the subject is one of the 1% of the population capable of absolute pitch or perfect pitch then a pitch above 2.5 kHz, even when presented with a reference, cannot be discriminated. It is worth noting that due to this, tones of frequency greater than 5 kHz cannot evoke a sensation of melody [4]. We can easily test out frequency discrimination in Matlab using the tonegen() function that was given in Section 2.7.1. We simply create and replay slightly different pure sinusoidal tones and see if there is any noticeable difference. In this case, let us use 1000 Hz and 1002 Hz tones of two seconds duration at 8 kHz: t1=tonegen(1000, 8000, 2); t2=tonegen(1002, 8000, 2); soundsc(t1, 8000); soundsc(t2, 8000); For a more scientific test we would preferably present a sequence of such tones in an unknown sequence and ask listeners to identify which are higher and which are lower. Over many repetitions, a score significantly greater than guesswork (50% correct) would indicate that the tones can be discriminated. 4.2.12 Pitch of complex tones A series of tones at 200, 400, 600, 800 Hz, . . . , evokes a pitch sensation of 200 Hz (as may be expected since this is the fundamental frequency in the set), however removal of the 200 Hz frequency does not affect the overall perceived tone. It does, however, affect the timbre or quality of the sound experienced. The perceived tone, when not actually present, is termed the residue [4]. This effect, which is known to musicians and composers, is very easy to demonstrate using Matlab. First we create four pure sinusoidal tones of duration two seconds at an 8 kHz sample rate. We will use a fundamental of 196 Hz (G 3 on the musical scale) and its multiples: t1=tonegen(196, 8000, 2); t2=tonegen(196*2, 8000, 2); t3=tonegen(196*3, 8000, 2); |
Ma'lumotlar bazasi mualliflik huquqi bilan himoyalangan ©fayllar.org 2024
ma'muriyatiga murojaat qiling
ma'muriyatiga murojaat qiling