Aspen is a unified accelerator for deep neural network (DNN)-based extended reality perception workloads. Aspen proposes a mixed-precision quantization scheme using the posit datatype to reduce memory usage while maintaining accuracy, a DNN accelerator for mixed-precision posit datatypes, and efficient data prefetching and data layout to minimize data reorganization. The Aspen system-on-chip has an Arm Cortex-M3 CPU, a mixed-precision posit-based DNN accelerator, and 4 megabytes of SRAM partitioned into eight 512 KB banks, connected through a 128-bit-wide interconnect.