Real-Time Food Intake Classification and Energy Expenditure Estimation on a Mobile Device
Obesity is a growing global health problem that has received increasing attention in recent years. It has been estimated that over 700 million people in the world are classified as obese. In the UK, the obese population has more than triple in the last 25 years. Obesity has been identified as an escalating global epidemic health problem and is found to be associated with many chronic diseases, including type 2 diabetes, cardiovascular diseases and cancer. Although there is well-publicised guidance on recommended daily calories intake, very seldom people will comply with such guideline as recording of calorie intake is time consuming and inaccurate, as methods for dietary and daily activity assessments mostly rely on questionnaires or self-reporting. New approaches have recently developed for the objective assessment of free-living food intake linking with daily activity patterns, increasing research in this direction is performed in recent years. For example, signal processing algorithms have been developed to detect and characterise food intake by capturing the sound generated during chewing and swallowing of food using an in-ear microphone. Wearable sensors have been used for objective monitoring of eating behaviour. With increasing capabilities of artificial intelligent, recent machine leaning approaches can accurately recognise patterns and classify images even with low power computing architectures. Several solutions that use smartphones or wearable devices have been proposed for managing dietary intake and monitoring energy expenditure to create an assistive calorie measurement system and help users to better manage their diet. We have recently introduced a deep learning image recognition approach for automatic food detection on a smartphone platform and released a mobile application, called FoodRec.
Comments