Electronic Nose Predicting Odor Pleasantness via Neural Network Mapping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current electronic noses (eNoses) primarily detect and discriminate between odors they have previously 'learned' and are unable to effectively predict the perceptual qualities of novel odors, particularly odorant pleasantness, which is a key aspect of human olfaction.
Innovation Solution
An eNose is tuned to a pleasantness scale using a pre-learned neural network that maps odor information from non-specific chemical sensors to an axis of odor pleasantness, allowing it to predict the pleasantness of novel odors by correlating sensor signals with human-assigned pleasantness scores.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If an eNose uses non-specific chemical sensors to detect odor patterns, then it can identify odor mixtures as a whole, but it cannot predict perceptual qualities of novel odors
Solution Approach 1:
The system performs preliminary action by collecting human pleasantness ratings for a comprehensive training set of odorants before deployment. This pre-learning phase enables the neural network to establish correlations between eNose sensor patterns and human perceptual responses, allowing the system to later predict qualities of novel odors without requiring real-time human input.
Solution Approach 2:
A neural network serves as an intermediary between the eNose sensor array and human perceptual responses. The neural network learns to map sensor response patterns to pleasantness ratings, bridging the gap between machine detection capabilities and human perceptual qualities. This intermediary enables the system to translate non-specific sensor data into meaningful perceptual predictions.
2Measurement precision
If an eNose is trained on a limited set of odorants, then it achieves significant prediction rates for those specific odors, but the results do not generalize to novel odorants
Solution Approach 1:
The system achieves universality by training the neural network on a diverse and comprehensive set of odorants covering multiple odor categories and chemical structures. This broad training enables the model to generalize its learned correlations to novel odorants across different chemical families, making the system universally applicable rather than specialized for specific odor types.
Solution Approach 2:
The system applies parameter changes by varying the composition, size, and diversity of the training set. By systematically changing the parameters of the training data (number of odorants, chemical diversity, concentration ranges), the neural network learns robust patterns that remain valid across different conditions and novel odorants, improving generalization capability.
3Measurement precision
If an eNose focuses on predicting discreet perceptual characteristics, then it achieves accurate predictions for those specific traits, but fails to capture the continuous nature of odor pleasantness
Solution Approach 1:
The system transitions from discrete categorical predictions to continuous dimensional scaling by implementing a pleasantness intensity axis. Instead of classifying odors into discrete pleasant/unpleasant categories, the neural network predicts continuous pleasantness values along a scaled dimension, capturing the nuanced gradations of human perceptual experience and avoiding oversimplification.
Data Source
AI summary
Apparatus and method for assessing odors, comprises an electronic nose, to be applied to an odor and to output a structure identifying the odor; a neural network which maps an extracted structure to a first location on a pre-learned axis of odor pleasantness; and an output for outputting an assessment of an applied odor based on said first location. The assessment may be a prediction of how pleasant a user will consider the odor.