Edge Sensor System for Activity Recognition

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing sensor systems for activity recognition, such as those described in US 2019/0103005 and US 2018/0306609, face challenges with user privacy, reliance on cloud connectivity, high energy consumption, and inefficiencies in processing data from multiple sensors, particularly audio data, which can lead to system interruptions and difficulties in adapting models to specific locations.

Innovation Solution

A sensor system with at least two sensors located in the same area, each equipped with a data processing unit and feature extraction unit, processes data locally to recognize primary activities without transmitting data to the cloud, allowing for robust and reliable activity recognition, energy efficiency, and adaptability to specific environments.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If data is continuously transmitted to the cloud for processing, then activity recognition can be achieved, but user privacy is compromised and energy consumption increases

Engineering Contradiction:
Improveactivity recognition accuracyVSAvoidenergy consumption
Core Design Contradiction:
ReliabilityVSUse of energy by moving object

Solution Approach 1:

The patent segments the data processing function by deploying edge computing nodes within the sensor system that perform local feature extraction and activity recognition, separating the data collection function (sensors) from the heavy processing function (edge nodes), thereby reducing cloud transmission requirements and energy consumption while maintaining recognition accuracy

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent introduces edge computing nodes as intermediary components between sensors and cloud infrastructure. These nodes process sensor data locally, extracting features and performing recognition before sending only essential results to the cloud, thus reducing energy consumption from continuous transmission while preserving activity recognition reliability

Inventive Principle:
Principle #24Intermediary (Mediator)

2Reliability

If multiple acoustic sensors are used for robust activity recognition, then reliability improves, but device complexity increases

Engineering Contradiction:
Improveactivity recognition reliabilityVSAvoidsensor system complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent merges the processing capabilities of multiple acoustic sensors by implementing a sensor fusion approach where edge computing nodes aggregate and correlate data from multiple sensors, performing unified activity recognition that leverages redundancy for improved reliability while presenting a simplified interface to users

Inventive Principle:
Principle #5Merging (Combining)

3Loss of energy

If all sensor data is processed locally, then user privacy is protected and energy consumption is reduced, but adaptability to specific locations decreases

Engineering Contradiction:
Improveenergy efficiencyVSAvoidlocal model adaptability
Core Design Contradiction:
Loss of energyVSAdaptability or versatility

Solution Approach 1:

The patent implements preliminary action by pre-training machine learning models at edge computing nodes with generic activity patterns before deployment. These pre-trained models provide baseline activity recognition capability locally, enabling immediate functionality with reduced energy consumption while maintaining adaptability through subsequent fine-tuning with local data

Inventive Principle:
Principle #10Preliminary action

4Productivity

If cloud connectivity is required for data processing, then comprehensive analysis is possible, but system reliability decreases due to connection dependencies

Engineering Contradiction:
Improvedata processing capabilityVSAvoidsystem reliability
Core Design Contradiction:
ProductivityVSReliability

Solution Approach 1:

The patent applies local quality by distributing processing intelligence to edge computing nodes deployed at local sensor locations. These nodes perform feature extraction and activity recognition independently using local data, ensuring system reliability during cloud disconnections while maintaining comprehensive analysis capability through localized machine learning models

Inventive Principle:
Principle #3Local quality

Data Source

PatentEP3767602A1Sensor system for activity recognition
Publication Date: 2021.01.20 NIKO NV
  • EP3767602A1 patent drawingFigure 1~2
  • EP3767602A1 patent drawingFigure 3
  • EP3767602A1 patent drawingFigure 4~5

AI summary

The present invention provides a sensor system (1) for activity recognition. The sensor system (1) comprises at least two sensors (S1, S2, ..., Sn) for capturing environmental data, a data processing unit (3) for each of the at least two sensors (S1, S2, ..., Sn) for processing the captured data, a feature extraction unit (6) for each of the at least two sensors (S1, S2, ..., Sn) for compacting the processed data by filtering out of the processed data information irrelevant for the activity, thereby obtaining activity relevant data, and a primary activity recognition unit (10) for, from the extracted relevant data, recognizing a primary activity. The feature extraction system and the primary activity recognition unit (10) are part of the sensor system (1), or in other words are located in the sensor system (1).