Neuromorphic Sensor Pipeline for Event-Driven AI Inference

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional CPUs are inefficient in processing machine learning tasks, particularly deep learning, due to high power consumption and limited processing capability, which is a challenge for battery-powered devices like smartphones.

Innovation Solution

Development of neuromorphic integrated circuits (ICs) that simulate neurons for parallel processing, utilizing bursts of electric current at non-uniform intervals, enabling efficient and powerful AI algorithms with up to 100x less energy consumption than GPUs and 280x less than digital CMOS solutions.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Power

If conventional CPUs are used for machine learning processing, then processing capability is provided, but power consumption is high

Engineering Contradiction:
Improvepower consumptionVSAvoidprocessing capability
Core Design Contradiction:
PowerVSProductivity

Solution Approach 1:

The patent replaces conventional von Neumann architecture CPUs with neuromorphic ICs that directly implement neural network computations using hardware circuits. This substitution enables efficient spiking neural network processing with event-driven computation, achieving both low power consumption and high processing capability for machine learning tasks.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The patent changes the fundamental operating parameters from clock-based sequential processing to event-driven parallel processing. By using spiking neural networks with temporal coding and asynchronous event propagation, the system achieves dynamic power adjustment and high throughput simultaneously, resolving the contradiction between power consumption and processing capability.

Inventive Principle:
Principle #35Parameter changes

2Speed

If conventional CPUs process machine learning tasks, then instructions are executed sequentially, but processing speed is limited

Engineering Contradiction:
Improveprocessing speedVSAvoidprocessing architecture
Core Design Contradiction:
SpeedVSDevice complexity

Solution Approach 1:

The patent segments the neural network processing into distributed spiking neurons and synapses across multiple neuromorphic cores. Each core processes independent neural computations in parallel, with event-driven communication between cores. This segmentation enables massive parallel processing speedup while maintaining manageable complexity through modular architecture.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent transitions from sequential time-based processing to parallel space-based processing. By implementing spatial parallelism across multiple neuromorphic cores handling different neural network layers or data streams simultaneously, the system achieves exponential speedup in machine learning inference while distributing computational complexity.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

3Use of energy by moving object

If neuromorphic ICs are used, then power consumption is reduced, but processing capability must be sufficient for AI algorithms

Engineering Contradiction:
Improveenergy consumptionVSAvoidprocessing capability
Core Design Contradiction:
Use of energy by moving objectVSProductivity

Solution Approach 1:

The patent implements periodic spiking activity driven by incoming sensor events rather than continuous clock operation. Neuromorphic ICs process information through discrete, event-triggered spikes that propagate through synapses only when needed, achieving both low energy consumption (computing only on events) and high processing capability (rapid event response).

Inventive Principle:
Principle #19Periodic action

Solution Approach 2:

The patent maintains continuous readiness of neuromorphic circuits in event-driven mode, where spiking neurons remain in a low-power idle state until triggered by sensor events. This continuity of useful action ensures the system is always prepared for rapid processing while consuming minimal energy during idle periods, balancing energy efficiency with processing capability.

Inventive Principle:
Principle #20Continuity of useful action

Data Source

PatentEP3830820B1Sensor-processing systems including neuromorphic processing modules and methods thereof
Publication Date: 2025.11.12 SYNTIANT
  • EP3830820B1 patent drawingFigure 1
  • EP3830820B1 patent drawingFigure 2
  • EP3830820B1 patent drawingFigure 3

AI summary

Disclosed is a sensor-processing system including, in some embodiments, a sensor, one or more sample pre-processing modules, one or more sample-processing modules, one or more neuromorphic integrated circuits ("ICs"), and a microcontroller. The one or more sample pre¬ processing modules are configured to process raw sensor data for use in the sensor-processing system. The one or more sample-processing modules are configured to process pre-processed sensor data including extracting features from the pre-processed sensor data. Each of the neuromorphic ICs includes at least one neural network configured to arrive at actionable decisions of the neural network from the features extracted from the pre-processed sensor data. The microcontroller includes a CPU along with memory including instructions for operating the sensor-processing system. In some embodiments, the sensor is a pulse-density modulation ("PDM") microphone, and the sensor-processing system is configured for keyword spotting. Also disclosed are methods of such a keyword spotting sensor-processing system.