Regression Model Predicts Sound Pleasantness
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for evaluating sound pleasantness, such as sound quality evaluation and noise, vibration, and harshness (NVH) analysis, are time-consuming and do not effectively account for human perception, particularly in industrial settings, where sounds can be unpleasant to some people but pleasant to others, leading to user fatigue and negative customer impacts.
Innovation Solution
A machine learning-based method and system that utilize regression prediction models trained on human ratings of sound qualities like loudness, tonality, and sharpness to predict the pleasantness of unrated sounds, allowing for efficient estimation and optimization of sound quality in devices and appliances.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a jury of listeners rates the pleasantness of sounds for each new product, then measurement precision of sound quality is improved, but loss of time increases significantly
Solution Approach 1:
The system performs preliminary action by collecting and storing sound quality measurements and pleasantness ratings from juries in advance during product development. These pre-collected data form a training dataset that enables the machine learning model to make rapid predictions without requiring real-time jury evaluations, thus resolving the time consumption issue while maintaining measurement precision.
Solution Approach 2:
The system creates a computational copy of the human jury's evaluation capability through a trained machine learning model. This model learns from historical jury ratings and sound measurements, then replicates the jury's pleasantness assessment function automatically. The copying principle allows the system to replace time-consuming human evaluations with fast computational predictions that preserve the original measurement precision.
2Measurement precision
If NVH analysis is performed in a laboratory setting with sensors, then measurement precision of sound qualities is improved, but adaptability to different products and user perceptions deteriorates
Solution Approach 1:
The system applies parameter changes by using machine learning to transform fixed laboratory measurement parameters into adaptive prediction parameters. The model learns optimal relationships between sound quality measurements (loudness, tonality, sharpness) and pleasantness ratings across different products and users. This allows the system to adapt to various products and individual user perceptions while maintaining the precision of laboratory-grade measurements through its training data.
Solution Approach 2:
The system achieves universality by creating a multi-functional evaluation platform that can assess sound pleasantness across different product types (electronic devices, appliances, vehicles) and account for individual user preferences. The machine learning model serves multiple functions: it performs NVH analysis, evaluates sound quality, predicts pleasantness ratings, and adapts to different products and users through its training data, replacing the need for separate laboratory analyses for each case.
3Device complexity
If machine learning models are trained with limited data, then device complexity is reduced, but measurement precision of predictions may deteriorate
Solution Approach 1:
The system applies partial action by selecting and using only the most relevant sound quality parameters (loudness, tonality, sharpness) from the full spectrum of possible acoustic measurements. This selective approach allows the model to achieve good prediction accuracy with limited training data by focusing on the key parameters that most strongly correlate with human pleasantness perception, rather than attempting to model all possible sound characteristics.
Data Source
AI summary
Machine learning is used to predict a pleasantness of a sound emitted from a device. A plurality of pleasantness ratings from human jurors are received, each pleasantness rating corresponding to a respective one of a plurality of sounds emitted by one or more devices. A microphone system detects a plurality of measurable sound qualities (e.g., loudness, tonality, sharpness, etc.) of these rated sounds. A regression prediction model is trained based on the jury pleasantness ratings and the corresponding measurable sound qualities. Then, the microphone system detects measurable sound qualities of an unrated sound that has not been rated by the jury. The trained regression prediction model is executed on the measurable sound quality of the unrated sound to yield a predicted pleasantness of the unrated sound.