Acoustic Feature Learning Device for Event Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing acoustic event classification techniques face reduced accuracy due to similarity between distinct acoustic signals, such as 'scream' and 'cheer', which results in similar acoustic features and decreased identification precision.
Innovation Solution
A learning device and method that extracts acoustic features, calculates language vectors from associated labels, and updates parameters to enhance similarity between acoustic features and language vectors, improving the identification accuracy of acoustic events by learning parameters that increase this similarity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If acoustic features are extracted using conventional parameters, then the extraction process is simple, but the identification accuracy of acoustic events is reduced when signals are similar
Solution Approach 1:
The patent changes the parameters used for acoustic feature extraction by learning optimal parameters that maximize the difference between acoustic features of different events. Instead of using fixed conventional parameters, the system learns parameters θ that transform acoustic signals into features where similar-sounding events (like scream and cheer) are differentiated more effectively, thereby improving identification accuracy.
Solution Approach 2:
The patent implements a feedback mechanism where the identification accuracy is used to guide parameter learning. The system calculates the difference between acoustic features of different events and uses this feedback to adjust and optimize the extraction parameters iteratively, allowing the system to learn from its performance and improve accuracy over time.
2Measurement precision
If acoustic features are extracted using learned parameters that maximize similarity difference, then identification accuracy improves, but the parameter extraction and learning process becomes more complex
Solution Approach 1:
The system performs self-service by automatically learning and optimizing its own extraction parameters without requiring manual tuning or complex external intervention. The parameter learning process is self-directed, where the system uses its own identification results to guide parameter optimization, reducing the need for external manufacturing complexity.
Solution Approach 2:
The patent applies preliminary action by pre-processing acoustic signals into acoustic patterns before feature extraction. This preliminary transformation prepares the data in a way that facilitates more effective parameter learning and improves the overall efficiency of the classification process, making the subsequent parameter optimization more manageable.
Data Source
AI summary
The acoustic feature extraction means 82 extracts an acoustic feature, using predetermined parameters, from an acoustic pattern obtained as a result of processing on an acoustic signal. The language vector calculation means 83 calculates a language vector from a given label that represents an attribute of a source of the acoustic signal and that is associated with the acoustic pattern. The similarity calculation means 84 calculates a similarity between the acoustic feature and the language vector. The parameter update means 85 learns parameters so that the similarity becomes larger, and updates the predetermined parameters to the parameters obtained by learning.