Analog in-memory compression and non-volatile storage for always-on sensing

Technology
In development
University
Scouted & summarized by Halo AI

An energy-efficient, always-on sensing architecture that utilizes analog in-memory compression and non-volatile storage to capture sparse data at nano-watt power levels. By eliminating the need for continuous power-intensive digital conversion, this system enables long-term, autonomous monitoring for rare-event data loggers and battery-constrained sensor nodes. The technology is available for sponsored replication, test-chip collaboration, or technology-transfer partnerships."}},technology_readiness_level:{citations:[1,2],confidence:

Overview

This analog always-on sensing architecture addresses the critical energy and information bottlenecks in edge computing by adopting a sense-now, reconstruct-later paradigm. By bypassing traditional power-intensive analog-to-digital conversion and digital storage, the system captures and compresses sparse events directly within a non-volatile memory array. This approach is highly effective for monitoring environments where infrequent, transient signals—such as seismic events, structural health data, or physiological activity—must be logged over long periods without frequent maintenance or high power consumption.

Technical specifications
  • Core Architecture: Employs an array of ultra-linear floating-gate (FG) injectors for non-volatile storage and in-memory time-domain multiply-accumulate operations.
  • Energy Efficiency: Operates at nano-watt power levels (approximately 530 nW measured in prototypes), significantly reducing consumption compared to traditional ADC-MCU-NVM data-logging pipelines.
  • Signal Processing: Uses a compact linear feedback shift register (LFSR) to generate pseudo-random measurement and encryption matrices in real time, enabling compressive sampling without storing large measurement matrices.
  • Versatility: Supports signal encryption and compression simultaneously. The hardware is fabricated in a standard CMOS process, facilitating potential scaling and integration.
  • Reconstruction: Signals are recovered offline from the accumulated statistics stored in the floating-gate array using optimized mathematical solvers.
Technology readiness level

4: The technology has been validated through fabricated hardware prototypes using standard CMOS processes. Experimental results demonstrate successful signal encoding, storage, and reconstruction under realistic conditions, with documented power consumption and performance metrics."},key_contact_email:{citations:[1],confidence:

Generated Aug 27, 2026. Please verify key information.

About University of Southern California

USC is a large, comprehensive private research university in Los Angeles, anchored by two urban campuses and an academic health system. Industry engages through applied research institutes such as the Information Sciences Institute, co-located engineering and clinical facilities, and flexible sponsored research contracting. Proximity to the region’s aerospace, media, and biotech corridors provides access to testbeds, talent, and pilot-scale partners across the greater L.A. innovation economy. Research is supported by competitive federal funding from NIH, NSF, and other major agencies alongside significant industry-sponsored awards. A dedicated technology transfer office supports IP strategy, licensing, corporate partnerships, and venture formation.

Halo home
Partner smarter. Move faster.
Get new partnering requests
delivered to your inbox.