Continuous wavelet transform (CWT) has been proven to be an effective tool in feature extraction of non-stationary bio-signals. Therefore, hardware implementation of CWT has been widely investigated in wearable sensor integrated with local intelligence algorithm. To realize the feature extraction from low-frequency bio-signals, the design method of low-frequency Gm-C wavelet filter used in wearable sensor has been proposed in this paper. To alleviate the power constraint by wearable device, the Leap-Frog multiple-loop feedback filter structure is employed, which has low circuit complexity and sensitivity. Also, the transconductor consisting of simple differential pair is employed as Gm cell. By using low level bias current and deep weak inversion, low transconductance can be achieved to realize low frequency operation. A sixth-order Gaussian wavelet filter is designed in 0.18 \(\upmu\)m CMOS process. Simulation results show that power consumption is only 392 pW at center frequency of 5.9 Hz, corresponding to dynamic range of 42.7 dB and figure-of-merit of 2.59 \(\times\) 10\(^{-13}\). Experiment result shows that the proposed wavelet filter can be used for accurate extraction of transient features in biomedical signal processing.

连续小波变换（CWT）已被证明是非平稳生物信号特征提取的有效工具。因此，在集成本地智能算法的可穿戴传感器中，CWT 的硬件实现得到了广泛的研究。为了实现低频生物信号的特征提取，本文提出了一种用于可穿戴传感器的低频Gm-C小波滤波器的设计方法。为了缓解可穿戴设备的功率限制，采用了Leap-Frog多回路反馈滤波器结构，电路复杂度和灵敏度低。此外，由简单差分对组成的跨导体用作 Gm 单元。通过使用低电平偏置电流和深度弱反相，可以实现低跨导，实现低频工作。\(\upmu\) m CMOS 工艺。仿真结果表明，中心频率为 5.9 Hz 时功耗仅为 392 pW，对应的动态范围为 42.7 dB，品质因数为 2.59 \(\times\) 10 \(^{-13}\)。实验结果表明，所提出的小波滤波器可用于生物医学信号处理中瞬态特征的准确提取。