MEMS and Sensors

Neuromorphic processors: The key to smarter sensing

03 February 2025
Innatera’s Spiking Neural Processor is targeted for all any use case where sensing is carried out continuously and can benefit from spatial and temporal patterns in the data stream. Source: Innatera

Neuromorphic processors, also known as brain-on-a-chip, are an emerging technology that powers neuromorphic computing, a technology that mimics a human brain’s neural structure.

Companies — from startups to tech giants — are developing neuromorphic processors for specific applications where systems can create efficient, robust systems able to learn and adapt like how a brain learns and adapts to various situations.

While potentially endless in tech applications, Innatera Nanosystems B.V., a spinoff from Dutch university TU Delft, is focused on using neuromorphic processors for smarter sensing. The company is focused on co-processors for small devices that operate on batteries using chips that process sensor data quickly but with little power.

What is neuromorphic computing?

Neuromorphic computing is a technology designed for systems inspired by the functionality of the human brain. The goal is to duplicate how the brain functions but for technology applications to make devices and systems more efficient, adaptive and intelligent.

The technology is particularly useful in tasks like:

  • Pattern recognition
  • Spatial data
  • Time-series analysis for robotics, edge computing and AI

The potential could create adaptable computing systems that are power efficient compared to conventional computing methods.

Neuromorphic processors rising

Because of the potential of this technology, semiconductor vendors are developing neuromorphic processors that contain neurons and synapses, the fundamental processing elements of the brain, using analog or digital circuits.

These devices mimic the brain’s information processing mechanisms. The chips can process data using sparse, asynchronous and event-driven neural networks for dealing with spatial- and time-series data.

The capabilities of such neural networks can additionally be augmented with processes that enable them to process analog data, perform online learning or adapt its functionality dynamically.

Sumeet Kumar is the CEO and co-founder of Innatera, developing neuromorphic processors.Sumeet Kumar is the CEO and co-founder of Innatera, developing neuromorphic processors.“Although neuromorphic computing has been researched for over 30 years in academia, the industry is only now discovering the applications where its efficiency advantages can be put to use — in sensing applications,” said Sumeet Kumar, CEO and co-founder of Innatera.

Processing sensor data on-chip

Innatera is developing what is claims is the world’s first ultra-low power neuromorphic microcontroller.

Called Spiking Neural Processor (SNP), the device allows for next-generation AI applications to push the sub-milliwatt power envelop. This allows for fast and efficient processing of sensor data on-device.

The SNP features on-chip RISC-V CPU co-located in its neuromorphic engine enabling raw data to be converted into insights with 10,000 times greater efficiency than traditional methods, Kumar said.

With that kind of power, even devices that run on small batteries will be able to run deep AI capabilities, he said.

How it would be used

Neuromorphic processors can be used for a range of applications. Innetara is focused on applications involving a range of sensors like:

  • Radar
  • Microphones
  • Inertial measurement units
  • Infrared sensors

Any use case where sensing is carried out continuously. Generally, applications using these types of chips involve finding spatial and temporal patterns in the data stream but other kinds of processing are also possible.

“We’ve demonstrated this with applications such as audio scene classification, our leading solution for robust human presence sensing with radars, as well as several other applications involving sensors that are increasingly finding a place in consumer and industrial devices,” Kumar said.

The next steps

In the next five to 10 years, neuromorphic computing, powered by neuromorphic processors, could potentially enable breakthroughs across the electronics landscape. This includes:

  • Energy-efficient AI at scale
  • Real-time learning and adaptation
  • Robotics that mimic humans
  • Neural implants for healthcare
  • Enhanced perception systems
  • Large-scale neuroscience simulations
  • Secure, decentralized AI

“We’re only starting to scratch the surface in terms of applications that can be enabled by neuromorphic computing,” Kumar said. “Our objective is to finally integrate closely with the sensor, enabling high-performance AI capabilities to be co-located with the sensor. This will lead to an ‘ambient intelligence’ that surrounds us, and effectively streamlines how sensor data is processed, and how AI is leveraged across the sensing value chain.”

To contact the author of this article, email PBrown@globalspec.com


Powered by CR4, the Engineering Community

Discussion – 0 comments

By posting a comment you confirm that you have read and accept our Posting Rules and Terms of Use.
Engineering Newsletter Signup
Get the GlobalSpec
Stay up to date on:
Features the top stories, latest news, charts, insights and more on the end-to-end electronics value chain.
Advertisement
Weekly Newsletter
Get news, research, and analysis
on the Electronics industry in your
inbox every week - for FREE
Sign up for our FREE eNewsletter
Advertisement
Find Free Electronics Datasheets
Advertisement