Wasserstein Metric for Noisy Sensor Data Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies face challenges in accurately detecting events, particularly in noisy data environments, such as those encountered in movement sensors or fluid systems, where noise accumulation complicates the identification of events like leaks.
Innovation Solution
The use of a data evaluation system that employs the Wasserstein metric to classify noisy data from sensors, allowing for the identification and classification of events by calculating distances between data sets and determining if they belong to the same category.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional machine learning algorithms are used to classify sensor data, then event detection is attempted, but noise accumulation in the data causes classification difficulty and reduced accuracy
Solution Approach 1:
The patent extracts and removes noise from sensor data before classification by comparing data patterns against known event signatures. The system identifies and eliminates noisy components that do not match characteristic event patterns, thereby improving classification accuracy without requiring complex noise filtering algorithms.
Solution Approach 2:
The patent introduces an intermediary classification layer that processes raw sensor data through pattern matching against predefined event templates. This intermediary step acts as a buffer between noisy raw data and final event detection, using the Wasserstein metric to measure similarity between observed patterns and known event signatures, thereby reducing noise impact.
2Reliability
If the Bucket Test method is used to detect leaks in fluid containers, then leak detection is attempted, but the method requires pumps and valve adjustments and can be inaccurate under certain conditions
Solution Approach 1:
The patent implements self-service leak detection by deploying sensors directly within the fluid container that autonomously monitor for leak conditions. The system automatically compares sensor readings against leak detection thresholds and generates alerts without requiring external intervention, pumps, or valve adjustments, thereby improving reliability while reducing operational complexity.
Solution Approach 2:
The patent replaces the mechanical Bucket Test system with electronic sensors and digital signal processing. Instead of using physical buckets, pumps, and valves to detect leaks, the system uses electronic sensors to monitor fluid levels and characteristics, substituting mechanical operations with electronic detection methods that are more reliable and easier to automate.
3Reliability
If cameras are used to monitor animal activity, then animal detection is attempted, but cameras are expensive and require great amounts of computer resources for processing
Solution Approach 1:
The patent employs inexpensive motion sensors instead of expensive camera systems for animal activity detection. These low-cost sensors consume minimal energy and provide sufficient detection capability for monitoring animal presence and movement patterns, replacing high-resource camera-based systems with energy-efficient alternative sensors.
Solution Approach 2:
The patent extracts only the essential detection functionality from complex camera systems by using dedicated motion sensors that detect animal activity without capturing video footage. This extraction approach removes the computationally intensive video processing requirement while retaining the core animal detection capability, thereby reducing energy consumption and computational resource requirements.
Data Source
AI summary
Devices, systems and methods for leak detection/identification are provided herein. Also provided are devices, systems and methods for monitoring and/or measuring fluid usage. In some aspects, a system comprising a sensor, a processing system, and a platform are provided. In some aspects, the sensor may be coupled to a spinning device. The sensor can be configured to detect fluid data, which can comprise, for example, displacement data of liquid and/or movement data associated with the liquid in a container and/or flow data associated with a flow of fluid in a conduit. The processing system can be coupled with the sensor and configured to communicate the fluid data. The platform can comprise an application communicatively coupled to one or more databases storing evaluation data (e.g., known pattern data) and configured to receive the fluid data and determine if there is a leak.