IEEE Trans Biomed Circuits Syst. 2007 Jun;1(2):128-35. doi: 10.1109/TBCAS.2007.907557.
Time-frequency domain signal processing of neural recordings, from high-density microelectrode arrays implanted in the cortex, is highly desired to ease the bandwidth bottleneck associated with data transfer to extra-cranial processing units. Because of its energy compactness features, discrete wavelet transform (DWT) has been shown to provide efficient data compression for neural records without compromising the information content. This paper describes an area-power minimized hardware implementation of the lifting scheme for multilevel, multichannel DWT with quantized filter coefficients and integer computation. Performance tradeoffs and key design decisions for implantable neuroprosthetics are presented. A 32-channel 4-level version of the circuit has been custom designed in 0.18-mum CMOS and occupies only 0.22 mm(2) area and consumes 76 muW of power, making it highly suitable for implantable neural interface applications requiring wireless data transfer.
从植入皮层的高密度微电极阵列中获取神经记录的时频域信号处理，对于缓解与颅外处理单元的数据传输相关的带宽瓶颈非常重要。由于离散小波变换（DWT）具有能量紧凑的特点，因此已被证明可以在不影响信息内容的情况下，为神经记录提供有效的数据压缩。本文描述了一种用于多电平、多通道 DWT 的提升方案的硬件实现，该方案具有量化滤波器系数和整数计算功能。为植入式神经假肢提出了性能权衡和关键设计决策。该电路的 32 通道 4 级版本已在 0.18 微米 CMOS 中定制设计，仅占用 0.22 平方毫米的面积，消耗 76 微瓦的功率，非常适合需要无线数据传输的植入式神经接口应用。
