Research On Autonomous Perception System Of Diet Composition Based On Service Robot's Hearing

As the problem of an aging population becomes more and more serious,more and more elderly people living alone need to be cared for,social robots for protecting and helping the elderly are now having an increasingly significant influence on human life.At the same time,with the continuous improvement of people's health awareness,dietary health has gradually become a growing concern.A scientific and rational diet is not only good for keeping healthy,but also can reduce a person's risk of developing cancer,diabetes,heart disease,and other age-related diseases.Although some social robotics products can establish personal health archives by employing regular question-and-answer conversations or wearable devices,they are unable to collect diet information automatically through robot vision or audition.Although the applicability and intelligence of existing social robots have been greatly improved,and health information can be acquired through regular dialogue and answering or wearable devices,there is not any existing social robot to automatically collect dietary information through its visual or auditory system,which ignores the significant impact of diet on human health.In order to improve the ability of social robots in health management,from the aspect of diet,which is closely related to people's life and has a great impact on human health,this paper uses the audition system of social robots to actively monitor the daily chat dialogue of users and acquire their dietary information independently,which provides a basis for users to adjust their unhealthy dietary structure,and at the same time provides a solution for mastering a person's dietary information in order to gain insight into the relationships between the occurrence of diseases and diets or to make an appropriate treatment prescription.The main research contents and innovation points of this research are as follows.1)The speech data processing method for different expression modes was designed to acquire the identity and contents of different speakers.Firstly,the voiceprint recognition method,speech segmentation clustering and speech recognition method used in this paper were summarized.Then,based on the features of different expression modes,a speech data processing framework for single-person expression and multi-person conversion was designed.According to the limited computing resource of the social robot platform,the voiceprint recognition method based on Mel frequency cepstrum coefficient and vector quantization algorithm was selected.At the same time,a voiceprint recognition method based on convolution neural network with less computation was designed.Finally,a voiceprint fragment clustering method based on sequence relations and identity features was proposed,which used the trained voiceprint recognition model to replace the similarity measurement method to cluster voice fragments.2)The automatic dietary composition perception algorithm based on social robot audition was proposed.Firstly,based on the national food classification method and the daily expression behavior patterns,the food lexicon was constructed by manual collection and web crawler technology.Then,taking Chinese word segmentation,part of speech and dependent syntax as characteristics,the reasoning definition of food acquisition based on the first-order predicate logic theory for single-person expression and multi-person conversion were proposed.According to the Chinese language expression pattern and the number of participants in the conversion,a dietary composition perception algorithm for single-person expression and multi-person conversion was designed.3)The algorithm proposed in this paper was implemented and integrated.The training and test data were collected to test the performance of the system,and the test results were further analyzed.The social robot platform was built and the algorithm was implemented using Python and C programming language.Firstly,the Mel frequency cepstrum coefficient of the optimal voiceprint recognition effect was obtained.Then,the improved evaluation index was proposed for the ambiguity and deceptive expression situation,and finally,the test results were analyzed and compared.The test results show that the automatic dietary composition perception system based on social robot audition can acquire of dietary information of different expression patterns in the home environment with good recognition effect and robustness.