Wyniki wyszukiwania - BazTech

1

Mixed signal correlator with extended DPCM characteristics

Lutowski Z.

Image Processing & Communications

|

2009

|

Vol. 14, no 1

35-42

EN

Correlation analysis (CA) of PCM signals permits to achieve high accuracy results at the expense of the use of multibit multiplication units. The use of mixed signal representation (PCM-DPCM) using quantization steps with multiplication factor equal to power of 2 permits to substitute multiplication with shift operations. To improve calculation accuracy addition and subtraction operations are introduced. General system structure and generation alghoritm are proposed.

2

The increase of DFT and DCT computation rate and accuracy with the use of parallel operations

Pogribny W., Drechny M.

Electronics and Telecommunications Quarterly

|

2007

|

Vol. 53, No 2

155-176

EN

The increase of DFT and DCT DFT (Discrete Fourier Transform) and its derivative DCT (Discrete Cosine Transform) are the transforms most often used in DSP (Digital Signal Processing), especially in data communications for signal compression [1, 2, 3, 4]. DFT and DCT algorithms have been modified and their rate and accuracy optimized for many years [3, 4]. Most of them are calculated in multibit PCM (Pulse Code Modulation) format. The differential DPCM (Differential Pulse Code Modulation) format, used in this work can be an alternative for PCM format applied in DFT and DCT. It ensures higher accuracy of computation with code word length shorter than PCM code word. When we modify DPCM format (Section 3) in such a way that the quantization steps are set of the numbers with a base 2 and exponent belonging to a natural numbers set, the multiplication operation rate, as one of the most often used operation in DSP, increases. It is possible because multiplication operations can be replaced with fast shift bit logical operations. The parallel combination of some MDPCM (Modified Differential Pulse Code Modulations) codes creates SDPCM (Synthesized Differential Pulse Code Modulation) code (Section 3), which has high computational accuracy, equal to the DPCM accuracy, however it does not require multiplications. In most cases, parallel computations lead to their rate increase in comparison to computation rate of sequentially operations. These calculations, apart from using appropriate and accurate algorithms require applying the systems which enable the effective work of the parallel methods. Thus, for this purpose the programmable FPGA devices (Fields Programmable Gates Array) have been the most commonly used recently. Their main advantages are high speed of operations, the possibility of programming every computational structure and their low price. In this work, apart from fast parallel DFT and DCT algorithms, we presented the structures of processing DFT and DCT systems (specialized processors) working in parallel way. The processing systems presented in Section 5 allow fast and accurate calculations without time-consuming multiplications. With reference to the article [5] presenting fast differential DCT algorithms, in this work the authors proposed another way of increasing the rate and accuracy of DCT computations, which consists in the modifications of a partV

3

Różnicowa DCT w zagadnieniach przetwarzania sygnałów

Pogribny W., Drechny M.

Kwartalnik Elektroniki i Telekomunikacji

|

2006

|

Vol. 52, z. 4

493-509

PL

Dyskretna transformata kosinusowa (DCT) jest szeroko stosowana do analizy i kompresji sygnałów. Klasyczne algorytmy DCT bazują na obliczeniach wykonywanych w wielobitowym formacie PCM (Pulse Code Modulation). Ograniczeniem zastosowania formatu PCM w DCT jest użycie wielobitowych operacji mnożenia, co często utrudnia jego wykorzystanie w systemach czasu rzeczywistego. Z drugiej strony, użycie niektórych niskobitowych rodzajów modulacji delta (DM) zapewnia taką samą dokładność obliczeń jak w przypadku użycia wielobitowego formatu PCM, lecz takie podejście nie zostało wystarczająco zbadane. Celem pracy było opracowanie i zbadanie efektywnych różnicowych algorytmów i struktur procesorów DCT pracujących w czasie rzeczywistym. W pracy tej przedstawiono nowe szybkie różnicowe algorytmy DCT. Szczególną uwagę skupiono na możliwości wyeliminowania w zaproponowanych algorytmach czasochłonnych operacji mnożenia i zastąpienia ich szybkimi operacjami bitowego przesunięcia SHIFT. Zastosowanie przesunięć umożliwiają kroki kwantowania będące potęgą liczby 2. Do zbadania dokładności zaproponowanych algorytmów wykorzystano sygnały szumu różowego oraz mowy. Zaproponowane zostały szybkie i efektywne struktury procesorów specjalizowanych realizujące opracowane algorytmy. Wyniki symulacji komputerowych potwierdziły efektywność opracowanych podejść.

EN

Discrete Cosine Transform (DCT) is widely used in analysis and signal compression. The classical DCT algorithms are based on calculations in multibit PCM (Pulse Code Modulation) format. The restriction of using PCM format in DCT is the application of multibit multiplication operations, which often limits the usage of PCM in real time systems. On the other hand, using some low bit kinds of delta modulation (DM) ensures the same calculation accuracy as in case of the multibit PCM format, however this approach has not been examined sufficiently yet. The purpose of the work was working out and studying the effective differential DCT algorithms and processors structures working in real time. In this work are presented new fast differential DCT algorithms. The particular attention was paid to the possibility of eliminating the multiplication operations in the proposed algorithms and replacing them by fast SHIFT operations. The application of shifts is done on the basis of the quantization steps, which are the power of number 2. To study the accuracy of proposed algorithms pink noise and voice signals were used. The authors proposed fast and effective structures of specialized processors realizing the worked out algorithms. The results of computer simulations confirm the efficiency of the presented approach.