Advanced semiconductor, visualisation and materials sciences specialist Nanoveu (ASX: NVU) has received notification that the United States Patent and Trademark Office (USPTO) has granted a US patent covering dedicated accelerator circuitry for non-maximum suppression in AI object detection.

The patent is assigned to NTU and is exclusively licensed to EMASS, Nanoveu's ultra-low-power semiconductor division.

Non-maximum suppression is the final stage in every AI object detection pipeline. After a detector proposes many overlapping candidate boxes for the same object, NMS sifts through them and keeps only the best match, discarding redundant detections. It is conventionally executed in software, where it consumes a disproportionate share of inference time and energy. The patented innovation performs the same function in purpose-built silicon, removing what the company describes as a structural bottleneck in edge vision.

In peer-reviewed research underlying the patent, a hardware implementation of the approach reported speedups of up to 75 times over a software implementation of the same algorithm, and 13.4 times over the prior state-of-the-art hardware accelerator, at comparable detection accuracy and full-HD input resolutions.

The patented mechanism is implemented as a dedicated hardware acceleration module in the BLE-enabled next-generation 16nm ECS-DoT chip, which entered fabrication at TSMC in January 2026. Moving to a 16nm process node is a meaningful step for a small-cap chip designer, as it places ECS-DoT in the same advanced-node territory used by mainstream edge AI silicon, where power efficiency per operation is a key differentiator.

Vision is a core edge AI workload, and efficient object detection is what makes always-on operation practical for battery-powered devices. Edge AI semiconductors compete largely on energy efficiency rather than raw throughput, because devices such as cameras and wearables must run inference continuously on tight power budgets rather than drawing on cloud data centres. The grant supports ECS-DoT's positioning in applications including smart cameras and doorbells, wearables such as smart glasses and AR/VR headsets, drones, and automotive sensing.

EMASS founder Dr Mohamed Sabry said object detection is where edge vision silicon wins or loses on power, and that non-maximum suppression is the stage conventionally left to software.

"This patent covers executing it in dedicated hardware instead, and that module is built into our next generation 16nm silicon. The exclusive licence puts that mechanism in EMASS's hands, and we intend to follow the same path from research to protected silicon across voice, motion and sensor fusion," Dr Sabry said.

The grant protects one workload within a substantially larger surface: ECS-DoT processes vision, audio and sensor data across the company's target verticals. For a company at Nanoveu's scale, an exclusive licence to granted US intellectual property can also strengthen its commercial positioning with larger semiconductor partners and customers, where freedom to operate on core pipeline functions is a standard diligence consideration. Nanoveu has previously stated its intention to file multiple additional patent families covering its drone energy optimisation and flight control technology, and intends to pursue further filings across voice, motion and sensor fusion. A near-term marker for investors to watch is ECS-DoT progressing through fabrication and moving toward expected delivery of working silicon.

NVU shares were steady at 5.8 cents, with a market capitalisation of $65.99 million, prior to markets opening.