Adaptive Integrate-and-Fire Time Encoding Machine With Quantization

An integrate-and-fire time-encoding machine (IF-TEM) is a power-effective asynchronous sampler that translates amplitude information into non-uniform time sequences. In this work, we propose a novel Adaptive IF-TEM (AIF-TEM) approach, which dynamically adapts the TEM bias and the induced Nyquist ratio in response to temporal amplitude and frequency variations of the input signal. We provide a comprehensive analysis of AIF-TEM’s oversampling and distortion properties. We also investigate the quantization process for AIF-TEM and analyze the corresponding mean squared error (MSE) bound. Our results show that AIF-TEM achieves significant improvements in rate-distortion performance compared to classical IF-TEM and traditional Nyquist (i.e., periodic) sampling methods for band-limited signals. In particular, AIF-TEM achieves at least a 12 dB reduction in reconstruction MSE under a fixed oversampling rate. When quantization is considered, AIF-TEM provides at least a 14 dB improvement in quantization MSE compared to IF-TEM. Furthermore, AIF-TEM achieves the same reconstruction accuracy using less than 30% of the total bits required by IF-TEM, highlighting the superior efficiency of our adaptive approach. Additionally, we introduce a dynamic quantization technique for AIF-TEM, which further improves performance by at least 10 dB compared to its classical quantization baseline.