[图文]Designandimplementationofasimpleinterneton the internetwifi无internet

周平文库

【图文】Designandimplementationofasimpleinternet

a simple design and implementation of led director with solar-po
a simple design and implementation of led director with solar-poDesignandimplementationofasimpleinterneton the internetwifi无internet

2017 2nd International Conference on Artificial Intelligence: Techniques and Applications (AITA 2017) ISBN: 978-1-60595-491-2Design and Implementation of a Simple Internet Violence Language Response System Based on NAO Robot Ni ZENG, Shi-jue ZHENG, Dong-dong LIU and Jun ZHOUSchool of Computer Science, Central China Normal University, Wuhan, China Keywords: NAO robot, Speech recognition and synthesis, Internet violence language.Abstract. With the progress of science and technology, robot technology has a rapid development and penetrated into the social production of all aspects. To achieve the natural communication between men and machine that voice interaction is the future of human- computer interaction mainstream. While network language constitutes an important part of people's daily life in the information age, some negative phenomena also arise—the abuse of the Internet violence language. In this paper, through the introduction of speech recognition and synthesis technology, the NAO robot is developed and embedded in the open artificial intelligence platform of IFLYTEK, which is based on the vocabulary of high frequency Internet violence and the corresponding quiz database is established at the same time. When people talk with the NAO containing the Internet violence language, the robot will prompt immediately and persuades them to use civilized language, which will standardize the network language environment, and further promote the construction of socialist spiritual civilization. Introduction Internet violence language is a special language phenomenon which is spread by a variety of media triggering a strong social concern or causing some damage to society. In the People's Network public opinion monitoring room released the "network of low-light language survey report" shows that "Nima" ranked first in the network of vulgar words list. The abuse of inappropriate language vocabulary under the Internet makes the social negative emotions, vulgar culture enlargement and words of intelligent age need to purify right now. In this paper, NAO uses the minimum mean square error estimation method (MMSE) for audio processing in the automatic detection of network vulgar vocabulary immediately and then makes a voice prompt response, breaking the traditional mode of human-computer interaction and reflecting the intelligent era under the emphasis on the construction of the network civilization environment based on speech platform API as a support. In this way, we can curb the network vulgar language, purify the virtual language communication platform better. Related Technology NAO Structure Function and Platform Development NAO, manufactured by Aldebaran Robotics, is a bipeded robot that combines a variety of hardware and software to support 90 minutes of development using the NAOqi development framework and DCM (Device Interaction Management System) to control the sensor. In the NAOqi OS above, we can call a large number of built-in library functions and procedures and achieve cross-platform, to achieve in Windows, Linux and MacOS can be based on this framework to prepare the program [1]. We write the python program on the Choregraphe platform, connect the NAO robot, and test the NAO robot's data changes through the real-time return of the computer. Speech Recognition and Synthesis Technology Speech recognition and synthesis technology to achieve the computer's intelligent, humane, is the information age under the great progress. The key technology in speech technology is divided into261

Designandimplementationofasimpleinternet

speech recognition technology ASR and speech synthesis technology TTS [4]. At present, the algorithms used are dynamic time regularization (DTW), HMM (Hidden Markov Model), and Gaussian Mixture Model (GMM, Gaussian Mixture Model)[5], in which HMM is the mostly used model for the characteristics of stochastic process statistics. Proposed Method Taking into account that the real experimental environment may have background noise and other factors, which will greatly affect the voice recognition rate, so we in this experiment taken by Ephraim and Malah proposed minimum mean square error estimation method (MMSE) [2,3]for voice Enhanced. One step preprocessing is required before the audio is transmitted to the IFLYTEK platform. Voice enhancement can reduce noise, improve voice clarity, and achieve higher speech quality. Its pure speech spectrum amplitude estimation value ̂݇ is=(1) For the analysis of the text through the BM algorithm to detect the match: If we have set the query class database in the vulgar language keywords and the matching index p is greater than a certain preset value, then determine the vulgar language question, again through the BM algorithm [10] Combine the corpus to match the match, the match is successful, the matching language to promote the language into speech output.k∞ ξk 1 e− x exp( ∫ dx) Rk 1 + ξk 2v xDesign of the Internet Violence Language Database We can classify high-frequency vocabularies in high-gloss vocabularies, form structures for vulgar vocabularies in different scenarios, write these high-frequency words into the database, and improve the content of the database in the later stages. The dialog form is shown in Figure 1.Figure 1. Dialog form.With the above method, we sum up the pre-collected network vulgar language and get the civilized example shown in Table 1In the database, firstly the keywords in the proposed problem are identified and recorded in the table. Secondly, match the search by the keyword and the vulgar language vocabulary. The corresponding civilized language tips are called immediately. Some examples of vulgar vocabulary and civilized hints are shown in Table 2.Table 1. Civilized dialogue form. d_id 1 d_type Medical care d_question How to register in the hospital How does the bank card report loss What is the first subway in the world d_answer Please go to the first floor lobby to show your ID card Please call the bank to freeze the phone immediately It’s the metropolitan metro of London, England in 1863. d_keyword1 hospital d_keyword2 register2Property securityBank cardloss3Life knowledgeThe firstsubway262

Designandimplementationofasimpleinternet

Table 2. Network vulgar word form. V_id 1 2 3 4 5 V_word Nima Cock silk Doubi Nimei Wokao V_tip Please use civilized language Please use the correct vocabulary You are not beautiful only civilized, but because of civilization is beautiful. Low-level dirty words vulgar piercing ears, civilized language elegant warm your heart. Everyone is responsible for creating a civilized society.Voice Call and Implementation The experimental environment is mainly for the NAO robot voice dialogue, where the network language vulgar language can be identified and responded immediately. The system resists the network vulgar language consciously and automatically. Through the microphone to the audio collection, the original analog signal data will be converted into digital signal, and then passed into the computer. In order to make the results of the match as accurate as possible, in the voice input stage will have those steps: pre-filtering, pre-emphasis, sampling and quantization, sub-frame, window, endpoint detection and other processes[7]. System speech recognition overall framework as Figure 2 shown below.Figure 2. System speech recognition framework.Digital signal, and then passed into the computer. In order to make the results of the match as accurate as possible, in the voice input stage will have those steps: pre-filtering, pre-emphasis, sampling and quantization, sub-frame, window, endpoint detection and other processes[7]. System speech recognition overall framework as Figure 3 shown below. As the NAO robot provides the Audio API for the developers to call, this paper uses the embedded IFLYTEK intelligent voice platform for voice recognition and synthesis. Call the voice module steps are as follows: First, the Choregraphe platform on the dialogue parameters are set. And the selected language type is English, through the transmission of parameters, the voice of the corresponding text in the database. Through the voice output, dang2 sets up after the completion of the instruction box of the connection and then operates on the platform, as shown in Figure 3 as shown in Figure 4.Figure 3. Select English.Figure 4. Dialogue parameters are set.263

Designandimplementationofasimpleinternet

The steps for voice acquisition and parameter transfer are as follows 1 Call the function RecognizerDialog (Context c, String "appid =" + appid :) Create a speech recognition control. OnResults (ArrayList results, boolean isLast) is used to get the recognition result, that is, the speech is converted to text. ArrayList is the result of speech recognition, where isLast is used to determine whether the recognized text is read. 2 Through the microphone input, the parameters of the information obtained through the network to the voice API analyzer. 3 Return the results into the temporary variables of the results, while the results of the corresponding regular expression to get the final analysis of the results. 4 When the system API identifies the corresponding voice, the converted text is passed to the database module to match and match the keyword. 5 If the match is successful in the network vocabulary vocabulary database, immediately prompt "please civilized language", call the function SynthesizerPlayer.createSynthesizerPlayer ().Figure 5. Python code operation.Figure 6. Choregraphe connection interface.Experimental Results In order to understand the recognition accuracy of NAO robot speech recognition on the vulgar vocabulary of the network, we have evaluated the recognition ability of it in the laboratory. First of all, we choose from the corpus 50 Chinese non-vulgar network words of the questioning sentence, defined as a civilized statement; and then in the 50 sentences of civilized sentences to add 1 to 2 vulgar network words, defined as a vulgar statement. Participate in the test of a total of 40 test subjects, including 20 men, 20 women, are using the more standard Mandarin recording. The specific test steps are as follows: 1. Each tester in turn read the random three civilized statements, each time the robot to answer after the record is expected to answer; 2. Each tester in turn read the random three vulgar statements, each time the robot to answer after the record is expected to answer; 3. Summarize the test results of all testers; 4. Statistics on the test results;Table 3. Test results. Testers sex Male Male Female Female Test statement type civilized statement vulgar statement civilized statement vulgar statement Completely correct quantity 57 57 55 58 Number of errors 2 3 3 1 Unrecognized quantity 1 0 2 1264

Designandimplementationofasimpleinternet

The test results are shown in the table. It can be seen that the accuracy rate of the grammatical recognition is 93.33%, the recognition rate of the vulgar statement is 95.83%, and the accuracy of the integrated speech recognition is 94.58%. From the feedback of the late testers, the cause of speech detection is a large part of the reason is difficult to voice recognition. On the whole, NAO robot speech recognition for most of the network vocabulary recognition accuracy is still very high. Conclusion In this paper, we achieve the system in which NAO robot can communicate with people through the application of speech recognition and synthesis technology in the Choregraphe platform for graphical programming that embedded in the IFLYTEK API interface. While realizing interesting intelligent voice dialogue between NAO robot and people, the Internet violence language can be identified automatically at the same time. It’s a great work to enhance artificial intelligence under the moral responsibility of consciousness, shape the modern network personality and guide the civilized way. Acknowledgements This work was supported by the key project China Language committee Research Project, Specific funding for education science research by self-determined research funds of CCNU from the colleges’ basic research and operation of MOE, No. ZDI135-13. References [1] Cai Hao. Network Development Based on NAO Service [J] .Journal of Jinan University, 2016. [2] Yang Long. Study on Several Single Channel Speech Enhancement Algorithms [J]. Science and [1] [3] Technology Horizon, 2015, 26: 155 + 222. [4] Cai Bin, Guo Ying, Li Hong-wei, Gong Cheng. An Improved MMSE Speech Enhancement Method [J]. Signal Processing, 2004, 01: 68-72. [5] Yu Yin-yu. Design and Development Based on Andrews Campus Service Dialogue System [D]. Fudan University, 2013: 12-13. [6] Zhang Jing-ya. Study on Chinese continuous digital speech recognition based on HMM [D]. Suzhou University, 2005:1-3. [7] Cai Hao. Network Development Based on NAO Service [J] .Journal of Jinan University, 2016. [8] Li Xiuzhen. Speech recognition algorithm and application technology research [D]. Chongqing University, 2010:1-6. [9] Network vulgar language blacklist released media title favorite with "Cock silk" [J]. Youth journalists, 2015, 17: 37:1-2. [10] Xiong Yuan. Network language violence research summary [J]. Young writers, 2014, 20: 79-80:1-2.265

Designandimplementationofasimpleinternet

Designandimplementationofasimpleinternet