|
Hi all,
Sorry, no Wavelet background, yet. Only wanna say thats a great pleasure to be into a mailing list where there are so knowledgeable people like Mr. Keith Larson (from who ive seen many ANs and example source codes), and Mr. Rulph Chassaing (from who i know he wrote a book on the C6x, although have not read it, yet), and probably many other distinguished people (because ive discovered this list very recently...) in the DSP fields. Also, many thanks to Mr. Dennis Paredes, who have already helped me off- list. Im an EE student, and with other students weve joined for starting a formal group in our School, basically a Lab for learning & developing DSP embedded applications. We only got a DSK C31 kit (by this first stage we think is enough!!!). So, my question is: Which road we should take in this amazing (but so big and confusing for beginners...) world of DSProcessors? A brief description of our actual learning stage, is the following. We already got: * A background on Signals & Systems and DSP basic theory. The course we took in our School was based mainly on the topics covered by the books: - "Signals & Systems", Oppenheim-Willsky - "Discrete-Time Signal Processing", Oppenheim-Shaffer * An extensive background and practise with 8-bit uCs. Actually, were getting: * Learning those "bridge" topics between DSProcessing theory and real- world DSProcessors (in our case, the TMS320C3X family), like: finite register length and its effects, FFT implementations of the DFT, floating- point representations and arithmetic, popular CODECs available, etc. * Learning about those C3x-specific topics, like: C3x assembly, DMA, etc. * Learning about the use of the C31 DSK. For these latter points were using mainly the (extensive and so useful!) literature available in TIs very complete web site. Also, we have requested (no response, yet) to Prentice Hall an evaluation sample of the following book, that seems to be very useful for us: - "A Digital Signal Processing Laboratory Using the TMS320C30" , H. Sorensen-J. Chen Any comment, suggestion, or pointer will be greatly appreciated!!! My wishes for the list getting more alive! Best Regards, S.- Sebastian Garcia School of Engineering University of Buenos Aires Buenos Aires, Argentina BTW: Please excuse my bad English language. > Hi Dennis > > I for one would not mind some more information with respect to > wavelets! > > My understanding of wavelets is that by correlating an input with a function or wavetable a weighting coefficient could be applied which more or less is the 'likeness' to the input. The input and 'likeness' are then differenced and the process repeated for other functions or wavetables until the residual is small. > > If this is true, then a DFT should also be a wavelet, albeit that the functions being correlated are sine and cosine waves. IOW, the DFTs complex output is a weighted series of strengths for the entire set of sines and cosines. If the DFTs output is then picked through for the strongest of its components, the original is more or less kept intact. > > But is this (DFT and the like) the best 'compressed' way to describe a signal (a window of samples)? I doubt it! > > If interested I created an adaptive filter (ANY2SIN.C) using a Sliding Fourier Transform (SFT) where I would use the SFT output to determine where the strongest harmonic was and then only reconstruct the output from that harmonics SFT bin. It would not be too dificult to add more harmonics and or pass out the strengths of each harmonic. > > Best regards, > Keith Larson > |