A Noise Filtering Algorithm for Event-Based Asynchronous Change Detection Image Sensors on TrueNorth and Its Implementation on TrueNorth

Abstract

Asynchronous event-based sensors, or “silicon retinae,” are a new class of vision sensors inspired by biological vision systems. The output of these sensors often contains a significant number of noise events along with the signal. Filtering these noise events is a common preprocessing step before using the data for tasks such as tracking and classification. This paper presents a novel spiking neural network-based approach to filtering noise events from data captured by an Asynchronous Time-based Image Sensor on a neuromorphic processor, the IBM TrueNorth Neurosynaptic System. The significant contribution of this work is that it demonstrates our proposed filtering algorithm outperforms the traditional nearest neighbor noise filter in achieving higher signal to noise ratio (~10 dB higher) and retaining the events related to signal (~3X more). In addition, for our envisioned application of object tracking and classification under some parameter settings, it can also generate some of the missing events in the spatial neighborhood of the signal for all classes of moving objects in the data which are unattainable using the nearest neighbor filter.

Spiking Neural Network Based Region Proposal Networks for Neuromorphic Vision Sensors

“…The importance of this layer is two-fold. Firstly, the output of DVS sensors often contains significant amount of noise [18] and the poor SNR affects the effectiveness of further processing of the events. With proper tuning of the refractory period, a sizeable fraction of these noisy events are eliminated without any considerable loss of signal and thereby, SNR is improved.…”

Distance Surface for Event-Based Optical Flow

“…Let Ψ in (4) denote the "ideal" event detection corresponding to the image edges. DVS suffers from considerably high noise (random events) along with the signal due to multiple factors such as electronic noise and sensor heat [51]. As such, the set of actual observed events Φ is a perturbed version of the ideal set of events Ψ in the following sense:…”

Low-Power Dynamic Object Detection and Classification With Freely Moving Event Cameras

“…The time thresholds for the nearest neighbor filter and the refractory filter are nominally set to be noise = 5 ms and ref = 1 ms, respectively, as suggested in Padala et al (2018). We used a FIFO buffer size of 5000 events for dynamically updating the count matrix as and when events are received.…”