System and Methods for Nutrition Monitoring

Summary

Diet is an important lifestyle and behavioral factor in self-management and prevention of many chronic diseases. Currently, technological approaches for nutrition monitoring either use computer vision techniques for image-based diet assessment or mobile apps for maintaining electronic food diaries. These approaches are lacking in accuracy, ease-of-use, and privacy and thus adherence to utilizing these technologies in clinical and real-world settings is quite limited.

Researchers at Washington State University’s School of Electrical Engineering and Computer Science have developed a system and associated algorithms and techniques for nutrition monitoring which employs speech recognition and natural language processing for measuring the nutrient information of food from spoken data. A user can essentially speak the name and portion size of the food they ate, and the speech is processed in real-time to compute the calorie intake. As we move into more speech-based user interfaces in mobile devices, this method allows user input in a more pervasive way while increasing the accuracy of calorie calculation.

This system also takes advantage of the United States Department of Agriculture’s (USDA) national database to compute the nutrient per serving calculations. This is an advantage over current calorie monitoring apps which utilize crowd-sourcing mechanisms to calculate caloric intake which can ultimately lead to many inconsistencies.

Advantages

•Uses a speech recognition system that performs speech transcription

•The system improves itself using automated prompting, thus the system generates prompts for the user to complete missing information (e.g. portion size)

•The system uses string matching to accurately match the food names and portion sizes in the nutritional database

•Robustness of the system has been validated in real-world, highly noisy settings

Patent Information:

Utility Patent Issued - US10078733B2 - System and methods for nutrition monitoring - Google Patents