Indoor Localization Using Binary Channel Measurement Compression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current indoor localization methods using Deep Neural Networks (DNNs) for channel measurements are resource-intensive, requiring high power, computational resources, and memory, especially as DNNs become more complex, and lack efficient solutions for low-power devices and processing channel measurements on-device.
Innovation Solution
A binary neural network-based approach that compresses channel measurements into a binary form, using bitwise operations and digital signal processing for pre-processing, allowing for efficient training and inference on low-power devices, reducing memory and computation requirements, and enabling on-device localization without dedicated hardware.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If Deep Neural Networks are used for indoor localization based on channel measurements, then localization accuracy is improved, but power consumption and computational resource requirements increase significantly
Solution Approach 1:
The patent transforms the continuous channel measurements into discrete binary features through quantization and thresholding operations. This parameter transformation reduces the computational complexity of neural network operations from floating-point arithmetic to simple binary comparisons, thereby significantly lowering power consumption while preserving localization accuracy
Solution Approach 2:
The patent extracts only the most relevant features from channel measurements by applying selective thresholding and binary encoding. Instead of processing all raw measurement data, the system extracts discriminative binary features that capture essential spatial information, reducing both computational load and energy requirements
2Measurement precision
If Deep Neural Networks are used for indoor localization, then localization accuracy is improved, but memory requirements and computational complexity increase
Solution Approach 1:
The patent changes the parameter representation from continuous floating-point values to discrete binary values (0 or 1). This transformation simplifies neural network computations to basic logical operations, dramatically reducing computational complexity and making the system feasible for deployment on resource-constrained devices
Solution Approach 2:
The patent employs simplified binary neural network models that can be efficiently trained and deployed on mobile devices. The binary nature of the model allows for compact representation and faster inference, effectively replacing complex floating-point models with lighter-weight alternatives
3Power
If channel measurements are processed using traditional methods, then computational resources are saved, but localization accuracy decreases
Solution Approach 1:
The patent applies binary quantization to channel measurement data, transforming continuous signals into discrete binary features. This parameter transformation enables the use of efficient binary neural network operations while maintaining the discriminative power needed for accurate localization, thus achieving both computational efficiency and high accuracy
4Reliability
If DNN models become more efficient and robust, then localization performance is improved, but energy and computational resource requirements increase
Solution Approach 1:
The patent quantizes channel measurement parameters into binary form, enabling the deployment of robust neural network models with reduced energy requirements. The binary representation maintains model reliability while dramatically lowering the energy cost of inference operations on mobile devices
Data Source
AI summary
An apparatus and method are provided for estimating a refined location in dependence on a plurality of measurements of one or more communication channels. The apparatus comprises one or more processors configured to: compress each channel measurement; process the compressed channel measurements using a neural network to form a plurality of intermediate location estimates; and process the intermediate location estimates to form the refined location.