Statistical Distribution Study of Dynamic Vision Sensor Data

This paper evaluates statistical distributions for modeling neuromorphic spike event data from Dynamic Vision Sensors (DVS) across diverse indoor and outdoor scenes. Using the DAVIS dataset, we applied and analyzed well-known distributions, including the Generalized Extreme Value (GEV) and Gamma distributions, by comparing empirical and fitted cumulative distribution functions (CDFs). Results indicate that while the GEV distribution consistently performs well for outdoor datasets, its accuracy varies across indoor scenes, highlighting the need for environment-specific modeling. These findings contribute to advancing the statistical representation and compression of neuromorphic vision sensor data.