Levesque Philippe, Sawan Mohamad
Polystim Neurotechnologies Laboratory, Ecole Polytechnique de Montréal, Québec, Canada.
IEEE Trans Ultrason Ferroelectr Freq Control. 2009 Aug;56(8):1654-65. doi: 10.1109/TUFFC.2009.1230.
A fully hardware-based real-time digital wideband quadrature demodulation processor based on the Hilbert transform is proposed to process ultrasound radio frequency signals. The presented architecture combines 2 finite impulse response (FIR) filters to process in-phase and quadrature signals and includes a piecewise linear approximation architecture that performs the required square root operations. The proposed implementation enables flexibility to support different transducers with its ability to load on-the-fly different FIR filter coefficient sets. The complexity and accuracy of the demodulator processor are analyzed with simulated RF data; a normalized residual sum-of-squares cost function is used for comparison with the Matlab Hilbert function. Three implementations are integrated into a hand-held ultrasound system for experimental accuracy and performance evaluation. Real-time images were acquired from a reference phantom, demonstrating the feasibility of using the presented architecture to perform real-time digital quadrature demodulation of ultrasonic signal echoes. Experimental results show that the implementation, using only 2942 slices and 3 dedicated digital multipliers of a low-cost and low-power field-programmable gate array (FPGA) is accurate relative to a comparable software- based system; axial and lateral resolution of 1 mm and 2 mm, respectively, were obtained with a 12-mm piezoelectric transducer without postprocessing. Because the processing and sampling rates are the same, high-frequency ultrasound signals can be processed as well. For a 15-frame-per-second display, the hand-held ultrasonic imaging-processing core (FPGA, memory) requires only 45 mW (dynamic) when using a 5-MHz single-element piezoelectric transducer.
提出了一种基于硬件的实时数字宽带正交解调处理器，该处理器基于希尔伯特变换来处理超声射频信号。所提出的架构结合了2个有限脉冲响应（FIR）滤波器来处理同相和正交信号，并包括一个执行所需平方根运算的分段线性近似架构。所提出的实现方式通过其动态加载不同FIR滤波器系数集的能力，能够灵活地支持不同的换能器。利用模拟射频数据对解调器处理器的复杂度和精度进行了分析；使用归一化残差平方和代价函数与Matlab希尔伯特函数进行比较。将三种实现方式集成到一个手持式超声系统中，用于实验精度和性能评估。从参考体模获取了实时图像，证明了使用所提出的架构对超声信号回波进行实时数字正交解调的可行性。实验结果表明，相对于类似的基于软件的系统，使用低成本、低功耗现场可编程门阵列（FPGA）的仅2942个切片和3个专用数字乘法器的实现方式是准确的；使用12毫米压电换能器时，无需后处理即可分别获得1毫米和2毫米的轴向和横向分辨率。由于处理速率和采样速率相同，高频超声信号也可以进行处理。对于每秒15帧的显示，当使用5兆赫兹的单元素压电换能器时，手持式超声成像处理核心（FPGA、内存）仅需要45毫瓦（动态）。
