Research Of Bimodal Co-development System Of Speech And Image Based On Developmental Network

With the deepening of computer technology, image processing, artificial intelligence, computer vision, physiology, psychology, cognitive science and many other related scientific research, speech recognition and image recognition has become a hot topic in the field of high-tech applications, is gradually into our lives. Currently, speech recognition and m image recognition technology used in almost navigation, Industrial robot control, mobile phone use, safety equipment and smart appliances, etc. and have a profound impact on people’s lifestyles. However, research on bimodal co-development of speech and image is very few; a bimodal co-development simulation system of speech and image is created based on developmental network in this thesis.Firstly, this thesis explains the specific structure and calculation processes of the basis developmental network, the learning algorithm of developmental network(lobe component analysis) and the influence of acetylcholine and noradrenalin to neural modulatory mechanism in detail, and describes in detail the model of neural modulatory system.Secondly, the thesis outlines the method of endpoint detection and the extraction of characteristic parameters, explains the specific structure and calculation processes of the system that combines developmental network and Mel Frequency Cepstral Coefficients(MFCC), and analyzes the influence of different parameters on the recognition rate of speech recognition.And then, this thesis takes the face recognition as an example to explain image recognition, explains the specific structure and calculation processes of face recognition that is based on developmental network. The experiment that the effect of different parameters on the recognition rate is done on face recognition against no background and the experiments that the effects of different number of iterations on the recognition rate are done on face recognition against complex background.Finally, a bimodal co-development simulation system of speech and image is created based on developmental network in this thesis. Voice and image library produced the paper made a brief description, and the specific structure and calculation processes of co-development system are described in detail. As can be seen from the experimental results, the average recognition rate of the co-development system is more than 98%, so we can get a conclusion that the bimodal co-development simulation system of speech and image that is based on developmental network can achieve good recognition results...