Popular:Hengli Industry LeTV Internet Café *ST longevity
Published on 2017-10-19 12:26:23 Oriental Fortune Network iPhone Edition
                        Heavy The world's first voiceprint recognition artificial intelligence TV listed Changhong set off an interactive revolution Pacific Home Network 10-19
                            Heavy The world's first voiceprint recognition artificial intelligence TV listed Changhong set off the interactive revolution
The Pacific OceanHome Network 10-19 09:51
Since the introduction of the TV for more than 80 years, the continuous innovation of display technology, from black and white to color, from 2D to 3D, from standard definition to ultra high definition, has become more and more beautiful, but it seems more and more difficult to keep the public on TV. In front of the machine, this is related to the outbreak of content sources in the Internet era, and rich and personalized content is presented through the network. In fact, not people don't like watching TV, but because most people don't have time to watch TV during the day, only the fragmentation time is left, so moving the small screen is more in line with people's needs, and the interaction is limited to switching channels with the remote control. Traditional TV began to become more alienated from users.
Changhong, which started to make TV voice interactions six years ago, brought the epoch-making artificial intelligence CHiQ TV 5K again on October 12th. This time, the tasting conference held in Chengdu will feature its far-field voice and voiceprint. The three core selling points of identification and fuzzy retrieval are fully displayed, especially the voiceprint recognition technology that Changhong applied for the first time in the artificial intelligence TV industry. It is also the key technology to open the gap with competitors in the same industry.
Voiceprint recognition gives TV the ability to identify family members
Voiceprint recognition technology, as its name suggests, can identify the specific identity characteristics of the person who issued the voice command, and then recommend audio and video programs according to each person's choice preferences and viewing history. The speech recognition function of other competing products still stays in the key issue of ignoring who issued the voice command, and only implements program recommendation based on historical browsing and semantic understanding. However, this is the least for home users. In the current technology of artificial intelligence TV, user behavior portraits are listed as one of the basic data, and the family members cannot be subdivided, which will greatly reduce the accuracy of user behavior analysis.
At the tasting session, CHiQ TV product manager Chen Keyu and another female guest played the role of family members. When the "I want to watch movies" instruction was issued, the recommended source of the CHiQ smart TV was not The same, and the interface also chooses different pattern colors according to voiceprint recognition. The interface and recommendation results of adult males, females and children correspond to their personality characteristics and viewing habits respectively, and when the guest opening of this TV has never been used CHiQ recognizes that this is a stranger, so the video is only recommended by default.
Chen Keyu introduced that Changhong applies the international frontier I-vector technology to voiceprint modeling, and at the same time, with the help of cloud computing platform, it establishes a unique voiceprint database for each family member, which can realize the family quickly and accurately in the home application environment. The identification of members, the recognition rate is over 90%.
It is understood that it is not easy to achieve such a good recognition rate. The Changhong voiceprint recognition project team encountered many difficulties in the development process. At the beginning of the project, the voiceprint data was one of the main difficulties faced, and it was impossible to build the model. Effective validation and training. The voiceprint recognition algorithm needs to accurately learn the difference between different speakers under the condition of enough training data. To this end, every member of the project team holds the remote control on time every day, constantly shouting "Changhong Xiaobai... I want to see... I want to see...", once and twice... countless times, collecting sounds day after day Grain sample.
With enough data, the design of voiceprint recognition algorithms is starting to become more important. Previously, the technical accumulation of the project team was mainly in the field of image recognition, and there was no experience in the field of speech related, so the design of the algorithm was particularly difficult. Professor Wu Wei, a core member of the project team, is in the United States. He meets twice a week through video conferences and discussions with domestic project team members. After three months of hard work, he finally completed the algorithm design, server deployment, and successful delivery. .
Fuzzy search and far-field speech make artificial intelligence more intimate
Many artificial intelligence TVs now use the built-in microphone of the remote control to collect user voice commands, which makes the consistency of voice interaction greatly compromised. The CHiQ5K is equipped with an integrated array microphone to achieve far-field voice collection within 5 meters, and can realize voice switching. From the time you turn on the TV to the completion of the program, you don't need a remote control for the whole process, only through voice control.
The computer can perform various tasks under precise instructions. The value of artificial intelligence is to identify natural dialogue and less precise instructions. The fuzzy search capability of Changhong CHiQ5 brings powerful correlation analysis and error correction capabilities. Even if the user is not accurate enough, it can present the video resources that the user wants.
Taking the fuzzy search function as an example, when the user clicks on the video, he accidentally presses the word "芈" in "芈月传" into a "half" character. At this time, the TV intelligent system can discriminate itself, and The keyword correction is performed according to the user's big data, so that the on-demand program is positioned on the "Yueyue Biography".
For the realization of this function, Chen Keyu said that "having such interactive functions requires the coordinated support of the technical system. Behind the artificial intelligence TV Q5K, there are basic data platforms, semantic platforms, user portrait platforms and media asset aggregation platforms. Support: TV has four algorithms systems: semantic understanding algorithm, retrieval algorithm, recommendation algorithm and data mining algorithm. There are also equipment control protocol, basic application management protocol, data query protocol and algorithm function protocol. Artificial intelligence with such technical system support TV Q5K has the characteristics of strong interaction, growth, consideration, and extensive cooperation with other devices."
Voiceprint recognition, far-field speech, fuzzy search, three characteristics that seem simple and refined, behind the accumulation of Changhong people's years of research and development results, Changhong's new generation of artificial intelligence TV Q5K, refreshing China's world of artificial intelligence A new level of technology, establishing a new benchmark for artificial intelligence TV, brings a revolutionary interactive experience to the traditional home entertainment center of television.
Published on 2017-10-19 15:04:53
                            d
Comment on this topic
The post is gone! How to do?
Author: You will not be publishedlog in |5 seconds registration Author:, welcome to leave a messagedrop out |Post a new topic
                Tip: All information, comments, etc. published by users in the community represent only personal opinions, and have nothing to do with the position of this website, and do not constitute any investment advice for you. Users should make their own decisions on securities investment and bear the corresponding risks based on their own independent judgment."Review of Self-discipline Management Commitment"
Heavy The world's first voiceprint recognition artificial intelligence TV listed Changhong set off the interactive revolution The Pacific OceanHome Network 10-19 09:51 Since the introduction of the TV for more than 80 years, the continuous innovation of display technology, from black and white to color, from 2D to 3D, from standard definition to ultra high definition, has become more and more beautiful, but it seems more and more difficult to keep the public on TV. In front of the machine, this is related to the outbreak of content sources in the Internet era, and rich and personalized content is presented through the network. In fact, not people don't like watching TV, but because most people don't have time to watch TV during the day, only the fragmentation time is left, so moving the small screen is more in line with people's needs, and the interaction is limited to switching channels with the remote control. Traditional TV began to become more alienated from users. Changhong, which started to make TV voice interactions six years ago, brought the epoch-making artificial intelligence CHiQ TV 5K again on October 12th. This time, the tasting conference held in Chengdu will feature its far-field voice and voiceprint. The three core selling points of identification and fuzzy retrieval are fully displayed, especially the voiceprint recognition technology that Changhong applied for the first time in the artificial intelligence TV industry. It is also the key technology to open the gap with competitors in the same industry. Voiceprint recognition gives TV the ability to identify family members Voiceprint recognition technology, as its name suggests, can identify the specific identity characteristics of the person who issued the voice command, and then recommend audio and video programs according to each person's choice preferences and viewing history. The speech recognition function of other competing products still stays in the key issue of ignoring who issued the voice command, and only implements program recommendation based on historical browsing and semantic understanding. However, this is the least for home users. In the current technology of artificial intelligence TV, user behavior portraits are listed as one of the basic data, and the family members cannot be subdivided, which will greatly reduce the accuracy of user behavior analysis. At the tasting session, CHiQ TV product manager Chen Keyu and another female guest played the role of family members. When the "I want to watch movies" instruction was issued, the recommended source of the CHiQ smart TV was not The same, and the interface also chooses different pattern colors according to voiceprint recognition. The interface and recommendation results of adult males, females and children correspond to their personality characteristics and viewing habits respectively, and when the guest opening of this TV has never been used CHiQ recognizes that this is a stranger, so the video is only recommended by default. Chen Keyu introduced that Changhong applies the international frontier I-vector technology to voiceprint modeling, and at the same time, with the help of cloud computing platform, it establishes a unique voiceprint database for each family member, which can realize the family quickly and accurately in the home application environment. The identification of members, the recognition rate is over 90%. It is understood that it is not easy to achieve such a good recognition rate. The Changhong voiceprint recognition project team encountered many difficulties in the development process. At the beginning of the project, the voiceprint data was one of the main difficulties faced, and it was impossible to build the model. Effective validation and training. The voiceprint recognition algorithm needs to accurately learn the difference between different speakers under the condition of enough training data. To this end, every member of the project team holds the remote control on time every day, constantly shouting "Changhong Xiaobai... I want to see... I want to see...", once and twice... countless times, collecting sounds day after day Grain sample. With enough data, the design of voiceprint recognition algorithms is starting to become more important. Previously, the technical accumulation of the project team was mainly in the field of image recognition, and there was no experience in the field of speech related, so the design of the algorithm was particularly difficult. Professor Wu Wei, a core member of the project team, is in the United States. He meets twice a week through video conferences and discussions with domestic project team members. After three months of hard work, he finally completed the algorithm design, server deployment, and successful delivery. . Fuzzy search and far-field speech make artificial intelligence more intimate Many artificial intelligence TVs now use the built-in microphone of the remote control to collect user voice commands, which makes the consistency of voice interaction greatly compromised. The CHiQ5K is equipped with an integrated array microphone to achieve far-field voice collection within 5 meters, and can realize voice switching. From the time you turn on the TV to the completion of the program, you don't need a remote control for the whole process, only through voice control. The computer can perform various tasks under precise instructions. The value of artificial intelligence is to identify natural dialogue and less precise instructions. The fuzzy search capability of Changhong CHiQ5 brings powerful correlation analysis and error correction capabilities. Even if the user is not accurate enough, it can present the video resources that the user wants. Taking the fuzzy search function as an example, when the user clicks on the video, he accidentally presses the word "芈" in "芈月传" into a "half" character. At this time, the TV intelligent system can discriminate itself, and The keyword correction is performed according to the user's big data, so that the on-demand program is positioned on the "Yueyue Biography". For the realization of this function, Chen Keyu said that "having such interactive functions requires the coordinated support of the technical system. Behind the artificial intelligence TV Q5K, there are basic data platforms, semantic platforms, user portrait platforms and media asset aggregation platforms. Support: TV has four algorithms systems: semantic understanding algorithm, retrieval algorithm, recommendation algorithm and data mining algorithm. There are also equipment control protocol, basic application management protocol, data query protocol and algorithm function protocol. Artificial intelligence with such technical system support TV Q5K has the characteristics of strong interaction, growth, consideration, and extensive cooperation with other devices." Voiceprint recognition, far-field speech, fuzzy search, three characteristics that seem simple and refined, behind the accumulation of Changhong people's years of research and development results, Changhong's new generation of artificial intelligence TV Q5K, refreshing China's world of artificial intelligence A new level of technology, establishing a new benchmark for artificial intelligence TV, brings a revolutionary interactive experience to the traditional home entertainment center of television.