Neuromorphic systems are bio-inspired and have the potential to break through the bottleneck of existing intelligent systems. This paper proposes a neuromorphic high-speed object recognition method based on DVS and SpiNNaker and implements a system in which an OR logic aggregation algorithm is used to acquire sufficient effective information and the asynchronous sparse computing mechanism of SNNs is exploited to reduce the computation. The experiment’s results show that the object detection rate of the designed system is more than 99% at the rotating speed of 900~2300 rpm; its response time is within $2.5 \mathrm{ms}$; and it requires 96.3% less computation than traditional recognition systems using the same scaled ANN.

The modeling procedure of current biological neuron models is hindered by either hyperparameter optimization or overparameterization, which limits their application to a variety of biologically realistic tasks. This article proposes a novel neuron model called the Regularized Spectral Spike Response Model (RSSRM) to address these issues. The selection of hyperparameters is avoided by the model structure and fitting strategy, while the number of parameters is constrained by regularization techniques. Twenty firing simulation experiments indicate the superiority of RSSRM. In particular, after pruning more than 99% of its parameters, RSSRM with 100 parameters achieves an RMSE of 5.632 in membrane potential prediction, a VRD of 47.219, and an F1-score of 0.95 in spike train forecasting with correct timing (±1.4 ms), which are 25%, 99%, 55%, and 24% better than the average of other neuron models with the same number of parameters in RMSE, VRD, F1-score, and correct timing, respectively. Moreover, RSSRM with 100 parameters achieves a memory use of 10 KB and a runtime of 1 ms during inference, which is more efficient than the Izhikevich model.