Sound Classification Apparatus Using Rule-Based Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing sound classification techniques rely heavily on the performance of the classification model, which requires a large amount of diverse training data to improve accuracy, making it difficult to achieve high classification accuracy without sufficient training data.
Innovation Solution
A sound classification apparatus and method that combines the output of a machine learning model with pre-registered information to classify sound data, allowing for improved accuracy independent of the classification model's performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If a classification model is used for sound classification, then automation is improved, but classification accuracy deteriorates when training data is insufficient
Solution Approach 1:
The patent segments the classification system into two independent parts: a machine learning classification model for automated classification, and a rule-based classification unit for handling specific conditions. This segmentation allows each part to excel at its designated task, with the rule-based unit compensating for the ML model's weaknesses when training data is insufficient.
Solution Approach 2:
The patent introduces a rule-based classification unit as an intermediary between the input sound data and the final classification result. This intermediary processes sounds that the ML model cannot accurately classify, thereby improving overall classification accuracy without compromising automation.
2Measurement precision
If more diverse training data is prepared to improve classification model performance, then classification accuracy is improved, but data preparation complexity increases
Solution Approach 1:
The patent segments the classification functionality between a trained ML model and a rule-based system. This allows the system to achieve high accuracy without requiring extensive diverse training data, as the rule-based unit handles cases where the ML model's training data is insufficient.
Solution Approach 2:
The patent performs preliminary classification using simple rules before resorting to the ML model or further processing. This preliminary action filters out cases that can be easily classified, reducing the burden on the ML model and decreasing the need for extensive training data.
3Device complexity
If only machine learning model output is used for classification, then device complexity is reduced, but classification accuracy deteriorates
Solution Approach 1:
The patent divides the classification system into an ML model component and a rule-based component. While this increases system complexity slightly, it dramatically improves classification accuracy by combining the strengths of both approaches.
Solution Approach 2:
The patent merges the ML model output with rule-based classification results to produce the final classification. This combination leverages both automated learning and deterministic rules, achieving superior accuracy compared to using either approach alone.
Data Source
AI summary
A sound classification apparatus includes: a learning model classification unit that inputs sound data to be classified into a machine learning model generated by machine learning using sound data and teacher data that serve as training data, and outputs a classification result using an output result from the machine learning model; a condition classification unit that classifies the sound data to be classified, based on information registered in advance, and outputs a classification result; and a sound classification unit 12 that classifies the sound data to be classified, based on the classification result of the learning model classification means and the classification result of the condition classification means.