Acoustic Impact Sorting With Machine Learning Feedback
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems for sorting objects with varying characteristics, such as electronic components or recyclable wastes, are costly, complex, and inefficient, particularly when implemented in pre-existing waste disposal chutes, and often result in errors due to the need for manual sorting or complex retrofitting.
Innovation Solution
A sorting system using acoustic detectors and machine learning algorithms to identify objects based on impact sounds, with feedback mechanisms to improve classification accuracy, allowing easy installation in existing chutes and reducing errors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If complex sensor sets or manual sorting are used, then sorting accuracy can be maintained, but system cost and time consumption increase significantly
Solution Approach 1:
The patent replaces complex mechanical sensor systems with an acoustic detection system that uses microphones to capture impact sounds. This substitution dramatically simplifies the device while maintaining sorting accuracy through acoustic signature analysis and machine learning classification.
Solution Approach 2:
The system creates acoustic copies or signatures of different object types by recording their impact sounds. These acoustic signatures serve as templates for classification, allowing the system to identify objects based on their acoustic fingerprints rather than requiring complex physical sensors.
2Measurement precision
If multiple detectors including cameras and metal detectors are used, then sorting precision improves, but system cost and complexity increase
Solution Approach 1:
The patent employs inexpensive acoustic sensors (microphones) instead of expensive detectors like cameras and metal detectors. This approach uses low-cost components that can be easily replaced or upgraded, significantly reducing system cost while maintaining effective sorting precision through acoustic analysis.
3Adaptability or versatility
If conventional sensors are used for waste sorting, then object classification can be achieved, but the system cannot be easily installed in pre-existing chutes
Solution Approach 1:
The acoustic detection system is designed to be universally applicable to various waste chute configurations. The system can be installed in pre-existing chutes without complex retrofitting, as acoustic sensors can be positioned to detect impact sounds from multiple object types regardless of chute geometry or material composition.
4Reliability
If manual sorting is performed, then sorting errors can be minimized through human judgment, but time consumption and labor costs increase
Solution Approach 1:
The system enables self-service sorting by using machine learning algorithms to automatically classify objects based on acoustic signatures. The system learns from training data and performs classification independently without human intervention, achieving both high reliability through accurate acoustic analysis and high productivity through automated rapid processing.
Solution Approach 2:
The system incorporates feedback mechanisms where classification results are continuously refined through machine learning. The system learns from correct and incorrect classifications, adjusting its acoustic signature matching algorithms to improve reliability over time while maintaining high sorting speeds through automated processing.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The system achieves high success rates (>70%) in sorting objects into multiple classes with reduced costs by leveraging cheap acoustic detectors and adaptive machine learning, minimizing errors through feedback training.
Implementation Method 1
at least one acoustic detector configured to capture acoustic data corresponding to the impact of a candidate object to sort on the impact body
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The present invention relates to a sorting system for sorting objects (100) belonging to at least two classes of objects having different acoustic signatures when impacting an impact body (50, 104), the system comprising: - at least one acoustic detector (60) configured to capture acoustic data corresponding to the impact of a candidate object (100) to sort on the impact body (50, 104), - a processor (66) for processing acoustic data to identify the class of the candidate object (100) executing a machine learning algorithm trained to identify the class to which a candidate object (100) belongs based on acoustic data of objects previously captured by the acoustic detector (60) and additional information describing said objects, and to generate a sorting information based on the identified class for the candidate object, - feedback means (80, 120) for providing to the processor (66) an additional information on the candidate object depending on the sorting information generated by the processor (66).