Odor to Color Transcription via Neural Network Analysis

The patent uses copying by creating virtual replicas of expert judge capabilities through artificial neural networks. Multiple neural network models are trained to replicate the decision-making patterns and chromatic assignment behaviors of human experts. These copied expert systems can be deployed in parallel to process multiple odor samples simultaneously, achieving high reliability through consensus algorithms while eliminating the logistical complexity of organizing and managing large panels of human judges.

The patent applies self-service by enabling the artificial neural network system to autonomously evaluate and assign chromatic profiles to new odors without requiring human judge intervention. The trained networks possess generalized knowledge from their training datasets, allowing them to independently classify novel odor samples by comparing them against learned patterns. This self-service capability provides immediate chromatic descriptions for new odors, eliminating the time delay associated with recruiting and training human judges for each new evaluation task.

The patent replaces the training process for human expert judges with an automated training process for artificial neural networks. The neural networks are trained using large, comprehensive datasets of standardized odor-chromatic associations, ensuring objective and consistent learning without the variability inherent in human training. This substitution eliminates the time and cost associated with recruiting, training, and certifying human expert judges while maintaining or improving measurement precision through the objectivity of machine learning models.