Implementation of Mellin transform using FPGA based embedded system platform
Implementation of Mellin transform using FPGA based embedded system platform
复制标题
基于FPGA的嵌入式系统平台实现Mellin变换
DOI：
10.1109/icpces.2012.6508091
复制
发表时间：
2012
期刊：
影响因子：
--
通讯作者：
K. C. Ray
中科院分区：
文献类型：
--
作者：
Kshitij;Shaik Abdul Rawoof;K. C. Ray
几十年来，梅林变换因其尺度不变性而引起了信号和图像处理等各种实时应用的关注，包括语音处理、模式识别和语音识别等。如今，所有这些应用都需要嵌入式平台来实时实现。由于Mellin变换是这些应用的组成部分之一，因此需要在嵌入式硬件平台上实现以满足实时计算的要求。据作者所知，还没有人提出Mellin变换在基于FPGA的嵌入式系统平台上的实现。近年来，基于FPGA（现场可编程门阵列）的实时嵌入式系统凭借其可重构性、灵活性、快速计算、设计上市时间短、成本低等特点，已经超越了微处理器、DSP等其他商用硬件平台。在此背景下，作者介绍了 Mellin 变换在嵌入式系统平台上的实现，该平台是使用 32 位 RISC 处理器 (TSK-3000A) 以及 FPGA (Xilinx Spartan XC3S1400AN) 芯片上的存储器、ADC 和 DAC 等定制外设开发的。分析了该嵌入式平台中使用缩放正弦信号作为输入的 Mellin 变换的性能，结果表明该变换具有尺度不变性，并且输出非常接近 MATLAB 仿真。
Since decades Mellin transform because of its scale invariant property has gained its attention to various real time applications such as signal and image processing including speech processing, pattern recognition and speech recognition etc. Today, all these applications demand embedded platform for real time implementation. Since the Mellin transform is one of the constituent in these applications, it needs to be implemented on embedded hardware platform to meet the real time computational requirements. To the knowledge of authors, no one have presented the implementation of Mellin transform on FPGA based embedded system platform. Recently, FPGA (Field Programmable Gate Array) based real time embedded system has taken its momentum over other commercially available hardware platforms such as microprocessors, DSP due to its features such as reconfigurability, flexibility, fast computation and less design time to market with low cost. In this context, authors have presented the implementation of Mellin transform on embedded system platform where this platform has been developed using 32-bit RISC processor (TSK-3000A) along with customized peripherals such as memory, ADC and DAC on a FPGA (Xilinx Spartan XC3S1400AN) chip. The performance of Mellin transform in this embedded platform using scaled sinusoidal signals as inputs have been analyzed which shows that the transform is scale invariant and output is very close to MATLAB simulation.