Edge Sensor Data Encoding for Private Low-Power Behavior Inference
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing monitoring technologies for vulnerable individuals or environments transmit raw sensor data, which is prone to interception and require significant computational power, leading to privacy and efficiency issues.
Innovation Solution
Edge devices encode sensor data using autoencoders and transmit encoded representations to a computing system, where prediction models are trained and redeployed, enhancing privacy, storage efficiency, and reducing power consumption.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If raw sensor data is transmitted to the computing system, then the system can perform comprehensive analysis, but the data becomes vulnerable to interception and requires significant computational power
Solution Approach 1:
The patent applies preliminary action by encoding sensor data at the edge device before transmission. The autoencoder compresses and encrypts the raw sensor data into encoded representations, so that when data reaches the computing system, it is already protected and compressed. This preliminary encoding action resolves the contradiction by establishing security and reducing computational burden before the data enters the main transmission and processing pipeline.
Solution Approach 2:
The patent segments the data processing functionality between the edge device and the computing system. The edge device performs initial encoding and compression using the autoencoder, while the computing system performs higher-level analysis. This segmentation allows the system to maintain security through distributed processing while reducing the computational power required at any single point, particularly at the edge device with limited resources.
2Loss of information
If raw sensor data is transmitted frequently, then the computing system receives up-to-date information, but the transmission consumes excessive power and bandwidth
Solution Approach 1:
The patent extracts only the essential information from raw sensor data by using the autoencoder to create compressed encoded representations. Instead of transmitting complete raw data sets frequently, the system transmits only the encoded essence of the data, which captures the critical information while occupying minimal bandwidth. This extraction principle resolves the contradiction by maintaining data freshness through efficient, low-power transmissions.
Solution Approach 2:
The patent changes the parameter of data representation from raw high-dimensional sensor data to compressed low-dimensional encoded data. The autoencoder transforms the data parameters, reducing the volume and complexity while preserving the essential information content. This parameter transformation allows frequent updates to be transmitted with minimal power consumption and bandwidth usage.
3Measurement precision
If the edge device processes and stores raw data, then accurate local predictions can be made, but the device requires significant memory and computational resources
Solution Approach 1:
The patent applies preliminary action by pre-training the autoencoder model offline and deploying it to the edge device. The encoding structure and compression algorithms are established in advance, allowing the edge device to perform accurate local predictions using the pre-configured model without requiring the device to store large amounts of raw training data. This preliminary preparation resolves the contradiction between prediction accuracy and memory requirements.
Solution Approach 2:
The patent uses copying by deploying a compressed version of the prediction model (the autoencoder) to the edge device. Instead of requiring the full training dataset and complex processing infrastructure at the edge, the system copies the essential encoded representation and model structure to the device. This copying approach enables accurate local predictions while keeping memory requirements manageable through model compression and optimization.
Data Source
AI summary
An example device is configured to encode first sensed data using a first encoder and to predict a first behavior based on the encoded first sensed data to create a first prediction using a first prediction model. The example device is configured to store the encoded first sensed data in the one or more memory units. The example device is configured to control the communication unit to transmit the encoded first sensed data in a first batch to a computing system. The example device is configured to receive, from the computing system via the communication unit, a second encoder, the second encoder being based at least in part on the encoded first sensed data. The example device is also configured to receive, from the computing system via the communication unit, a second prediction model, the second prediction model being based at least in part on the encoded first sensed data.