Architectures for Multiple Transforms for Signal Processing Applications

Abstract

Discrete transforms play major role in digital signal processing applications for processing various types of signals such as speech, image, video, radar, seismic and biomedical signals. Transforms are broadly classified as sinusoidal transforms and non-sinusoidal transforms based on the basis function used for representation. Software and hardware based approaches are the two main platforms for transform implementation. Hardware approach essentially requires design and development of new algorithms and architectures. Architectures for transforms are evaluated based on the performance metrics such as area, power and computation time (or throughput). Few specialized applications need multiple sinusoidal and non-sinusoidal transforms to improve the system performance. Standalone architectures for each transform demand higher area and power. Hence, to cater to these demands, it is advantageous to design a hybrid architecture supporting multiple transforms with support for scalability. There has been substantial development over the past few decades in design of efficient algorithms and architectures for standalone and hybrid architectures. However, architectures for transforms with sinusoidal basis and non-sinusoidal basis are an emerging research topic. This thesis is aimed at design, simulation, performance evaluation and implementation of standalone and hybrid architectures to support sinusoidal and non-sinusoidal transforms. Discrete Fourier Transform (DFT), Discrete Cosine Transform (DCT), Discrete Sine Transform (DST), Discrete Hartley Transform (DHT) and Discrete Wavelet Transform (DWT) are chosen due to their suitability in many applications. Cyclic convolution based approach is used to design an optimized architecture for 1-D DCT with almost 50% reduction in hardware resources. The method is extended to design an architecture for 1-D DFT and the commonality at architecture level is explored. Convolution based Filter bank approach and Lifting approach are the two main design techniques available for DWT. ...