Multimedia Systems & InterfacesImagesAffine TransformsPowerPoint PresentationSlide 5Slide 6Slide 7Slide 8Slide 9Slide 10Slide 11Slide 12Slide 13Slide 14Slide 15Slide 16Slide 17Slide 18Slide 19Slide 20Slide 21Slide 22Slide 23Slide 24Slide 25Slide 26Slide 27Slide 28SamplingTime Series - Audio ExampleAudio ExampleSlide 32Slide 33Slide 34Slide 35Slide 36Slide 37Slide 38Slide 39Slide 40Slide 41Slide 42Slide 43Slide 44AssignmentMultimedia Systems &InterfacesSlides adapted from material created by Chris Wren and Paris SmaragdisImages•Transforms•DFT, Sampling•Filter intro•AssignmentAffine TransformsSamplingTime Series - Audio ExampleAudio ExampleReal signal comparison0 0.5 1 1.5 2 2.5 3x 104-1-0.500.51Time domainTimeFrequency0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.602000400060008000•Time domain–We can see the events–We don’t know how they sound like though!•Spectrum–We can see a lot of bass and few middle freqs–But where in time are they?•Spectrogram–We can “see” each individual sound–And we know how it sounds like!0 100 200 300 400 500 60000.511.52Frequency domain0 10 20 30 40 50 60 70 80 90 100-1-0.500.51Signal representation - Quantization•Converting a pressure to a digital number is called digitization or quantization•We need adequate resolution to represent the pressure measurements–Precision is measured in bits–Possible noise with soft sounds–Possible distortion with loud sounds•Bit precision (headroom)–8-bit = 48dB, poor–12-bits = 72dB, maybe ok–16-bits = 96dB, good–24-bits = 144dB, little too much–Our ears deal with about 120dB•CDs, DVDs, etc are usually 16-bit–Pro-music machines are 24-bit•Sophisticated audio processing is in floating-point where digitization issues are mostly moot3-bit4-bit8-bit12-bitSoft sound problems (hiss)Loud sound problems (clipping)x140x14x3.5x1.75ClippingMisrepresentationCauseof hissQuantization in imagesAliasing•Low sample rates result in aliasing–High frequencies are misrepresented–Frequency f1 will become (sample rate – f1 )–In video also when you see wheels go backwardsAliasing examplesTimeFrequency0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.900.511.52x 104TimeFrequency0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.90200040006000800010000TimeFrequency0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9010002000300040005000Sinusoid sweeping from 0Hz to 20kHz44kHz SR, is ok 22kHz SR, aliasing! 11kHz SR, double aliasing!On real soundsat 44kHzat 22kHzat 11kHzat 5kHzat 4kHzat 3kHzOn videoOn imagesFilters•Often we want to change the character of a sound–E.g. a stereo EQ•This is accomplished using filters•Filters manipulate the spectrum of a sound–Boost some frequencies–Suppress others–Or vice versa …TimeFrequency0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.602000400060008000Filter types by function•Lowpass–Allows only low frequencies through•Highpass–Allows only high frequencies through•Bandpass–Allows only a band of frequencies through•Band-reject/stop–Allows everything but a band of frequencies through•Custom–Does some arbitrary selectionTimeFrequency0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.602000400060008000TimeFrequency0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.602000400060008000TimeFrequency0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.602000400060008000TimeFrequency0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.602000400060008000Band-reject BandpassLowpass HighpassBand-rejectAssignment1. Load an image of your choosing into matlab.2. Extract the red, green, blue channels of this image.3. Extract the luminance from this image.4. Explore different variations of this image on your own.5. Rotate the image 45 degrees.Create a document showing 1, 2, 3, one of your experiments from 4, and 5. Include all of your code in this document.Email this document (preferably in pdf form) to [email protected] by midnight before class on
View Full Document