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
61 Figure 4.2 Equal loudness contours (constructed from examination of Fletcher–Munson data). conducted by many researchers over the many years, which allow a model of the effects of psychoacoustics to be built up. In the subsections that follow, several interesting phenomena are described that have a direct impact on the subject, and which will form the foundations of many of the techniques described in Chapter 7. 4.2.1 Equal loudness contours Human subjects do not always judge tones of different frequency, but identical ampli- tude, to be equal in loudness [5,7]. Put another way, a 1 kHz signal at 40 dB SPL would sound louder than a 15 kHz signal at the same amplitude. Actually it is not surprising since 15 kHz is near the limit of hearing. After long and careful research, scientists identified ‘equal loudness contours’ in the 1940s that model this behaviour. These contours are plotted in Figure 4.2, and show the curves identified in phons, where a curve of n phons follows the contour of different frequency amplitudes that sound equally loud to a subject as an n dB SPL tone at 1 kHz. The lowest curve (the dashed line) indicates the threshold of hearing. The effect of equal loudness should be relatively easy to test in Matlab, but one should beware that dependent upon the frequency response of the audio loudspeakers or headphones being used, some sounds will naturally be reproduced with higher amplitude than others. Assuming a good pair of headphones with relatively flat frequency response is available, let us create three pure tones in Matlab using the tonegen() function from Section 2.7.1, and listen to them: lo= tonegen(250, 441000, 2); mi= tonegen(1200, 441000, 2); 62 Hearing hi= tonegen(11000, 441000, 2); soundsc(lo, 441000); soundsc(mi, 441000); soundsc(hi, 441000); Based on the loudness curves, the 1.2 kHz middle tone should be perceived as being louder than either the 250 Hz lower tone or the very high 11 kHz sound (which may not even be audible at all for many people). We have used a 44.1 kHz sample rate in this case in order to faithfully reproduce the 11 kHz sinewave. It should be noted that the equal loudness contours derive from averaged tests, and apply to single tones only: there may be substantial differences from person-to-person, especially in cases of people who listened to too much rock music when young (suffering from hearing damage), but in the experience of the author, the curves do describe the ensemble of human responses fairly accurately. Download 2.66 Mb. Do'stlaringiz bilan baham: |
Ma'lumotlar bazasi mualliflik huquqi bilan himoyalangan ©fayllar.org 2024
ma'muriyatiga murojaat qiling
ma'muriyatiga murojaat qiling