Implementation of Cosine Modulated Digital Filter Bank on Processor with ARM Architecture

This paper presents a class of multichannel cosine-modulated filter banks (CMFB) of analysis based on the modulation effect with a fast discrete-cosine transformation of the fourth type (DCT-IV), which is calculated using the fast Fourier transform. As a prototype filter, a low-frequency filter with a finite pulse characteristic was used, frequency-shifted copies of which were made using an effective technology for polyphase representation of the filter Bank. The comparison of the number of arithmetic operations performed by digital down converter (DDC) based on cascade integral-comb (CIC) and CMFB based on different number of channels is given. A software description of the CMFB algorithm is presented in the form of block diagrams describing the capabilities of the Opencl and clfft software libraries for implementing the DCT-IV filter Bank and modulation algorithm on a GPU. The obtained algorithm was tested on an ARM family processor and a mali GPU with a table with sample rate for different number of channels with the maximum load of the graphics processor (GPU) and the minimum load of the Central processor (CPU).

TIGHT FRAME CHARACTERIZATION OF MULTIWAVELET VECTOR FUNCTIONS IN TERMS OF THE POLYPHASE MATRIX

The extension principles play an important role in characterizing and constructing of wavelet frames. The common extension principles, the unitary extension principle (UEP) or the oblique extension principle (OEP), are based on the unitarity of the modulation matrix. In this paper, we state the UEP and OEP for refinable function vectors in the polyphase representation. Finally, we apply our results to directional wavelets on triangles which we have constructed in a previous work. We will show that the wavelet system generates a tight frame for L2(ℝ2).