Applied Speech and Audio Processing: With matlab examples
Download 2.66 Mb. Pdf ko'rish
|
Applied Speech and Audio Processing With MATLAB Examples ( PDFDrive )
4.5. Auditory scene analysis
83 To really show that this interpretation is not simply a side-effect of the small gap that was inserted, the reader is encouraged to repeat the experiment, but modulate bm and cm in the same way as am, namely: am=mod1.*(sa+sa2); bm=mod1.*(sb+sb2); cm=mod1.*(sc+sc2); With all subsequent steps repeated as previously, the listener will hear, instead of a discordant mixture of three notes, a warbling single chord. This is to be expected since now each of the three harmonic notes shares a common fate (a common modulation). 4.5.4 Good continuation Generally, sounds in nature do not start, stop or switch frequency instantaneously. There will normally be some amplitude attack at the beginning (ramping up from zero) and decay (ramping down to zero) at the end of a sound. Similarly, sounds in nature normally do not just switch frequency instantaneously, they slide either quickly or slowly. Both of these properties are due to the physical methods of creation of such sounds. Changes in frequency are caused by changes in the shape or geometry of the produc- ing object, such changes would be driven by muscles in humans and animals, which move relatively slowly and gradually in audio terms, so that at least some intermediate frequencies will be present between two extremes. Similarly, sounds are conveyed by the movement of air, and air movements can be sharp but not completely instantaneous. They must at first build up, and later die away. So when presented by complex sets of sound, the human brain tends to classify sounds that are connected in some way by these gradual changes as coming from a single source. By contrast, sounds which do not blend together are more likely to be regarded by the brain as coming from separate sources. Researchers have termed this effect the good continuation of sounds, although it could perhaps better be described as being the connectedness of individual sounds. Musical instruments are similar. In particular their notes always exhibit some form of amplitude attack and decay, however notes do not always glide into one another. Computers provide us the ability to generate almost any arbitrary set of sounds, with whatever features we may require. Computer generated sounds can start, stop and switch frequency instantaneously if we wish, or could be tailored to show some connectedness. In the following Matlab code, we will aim to demonstrate the effect of the good continuation property by constructing two sequences of sound. The first, shown in Figure 4.9(a), has two alternating notes joined by a sliding frequency, or glissando, while the second, in plot (b), presents the same two notes without any glissando. Listeners of both sequences should consider that the first could conceivably be produced by a single source which slides upwards and downwards in frequency. The second, by contrast, is more likely to be interpreted as two separate, single frequency sound sources that alternate in producing a sound. |
Ma'lumotlar bazasi mualliflik huquqi bilan himoyalangan ©fayllar.org 2024
ma'muriyatiga murojaat qiling
ma'muriyatiga murojaat qiling