Food intake calorie prediction using generalized regression neural network

Many devices have been proposed to monitor the calorie intake and eating behaviors. These wearable devices uses various sensing modalities, such as acoustic, visual, inertial, EEG (electroglottography), EMG (electromyography), capacitive and piezoelectric sensors. In this paper, Generalized Regr...

وصف كامل

محفوظ في:
التفاصيل البيبلوغرافية
المؤلفون الرئيسيون: Gunawan, Teddy Surya, Kartiwi, Mira, Abdul Malik, Noreha, Ismail, Nanang
التنسيق: Conference or Workshop Item
اللغة:English
English
منشور في: IEEE 2019
الموضوعات:
الوصول للمادة أونلاين:http://irep.iium.edu.my/72546/1/72546%20Food%20Intake%20Calorie%20Prediction.pdf
http://irep.iium.edu.my/72546/2/72546%20Food%20Intake%20Calorie%20Prediction%20SCOPUS.pdf
http://irep.iium.edu.my/72546/
https://ieeexplore.ieee.org/document/8688787
الوسوم: إضافة وسم
لا توجد وسوم, كن أول من يضع وسما على هذه التسجيلة!
الوصف
الملخص:Many devices have been proposed to monitor the calorie intake and eating behaviors. These wearable devices uses various sensing modalities, such as acoustic, visual, inertial, EEG (electroglottography), EMG (electromyography), capacitive and piezoelectric sensors. In this paper, Generalized Regression Neural Network (GRNN) will be utilized to predict the food intake calorie from the input of digital image. GRNN was utilized due its fast training compared to standard feedforward networks. The food image database comprises of 568 food including sweet, savory, processed, whole foods, and beverages. The calorie has the ranged from 0 kcal (plain water) to 11830 (roasted goose) with median 235.5 kcal. The optimum spread parameter for GRNN was found to be 0.46 when the 568 images was distributed randomly, i.e. 80% training and 20% testing. Due to very large variation of the calorie needs to be predicted, GRNN has rather large prediction error. This could be alleviated using more training data, use other features like texture and segmentation, or deep neural network.