Search

KR-102962201-B1 - ARTIFICIAL INTELLIGENCE SERVER

KR102962201B1KR 102962201 B1KR102962201 B1KR 102962201B1KR-102962201-B1

Abstract

An artificial intelligence server is disclosed. An artificial intelligence server according to an embodiment of the present invention includes a communication unit that communicates with one or more electronic devices, and a processor that receives input data from a specific electronic device, applies personalized information corresponding to the specific electronic device to a recognition model, inputs the input data to the recognition model to which the personalized information is applied to obtain a final result value, and transmits the final result value to the specific electronic device.

Inventors

  • 한종우

Assignees

  • 엘지전자 주식회사

Dates

Publication Date
20260508
Application Date
20190902
Priority Date
20190711

Claims (13)

  1. A communication unit that communicates with multiple electronic devices; A memory storing a common model including batch normalization parameters that are applied equally to each of the plurality of electronic devices, and a personalized model including batch normalization parameters that are applied differently to each of the plurality of electronic devices and varying the mean and variance corresponding to each of the plurality of layers through the batch normalization parameters; and A processor comprising: receiving input data from a specific electronic device among a plurality of electronic devices; inputting the input data into a common model to obtain an intermediate result value applicable to the plurality of electronic devices; receiving personalized information including a unique batch normalization parameter of the specific electronic device from the specific electronic device; inputting the personalized information including the unique batch normalization parameter into a personalized model to change the batch normalization parameter of the personalized model to the unique batch normalization parameter of the specific electronic device; inputting the intermediate result value into the personalized model to obtain a final result value applicable only to the specific electronic device; and transmitting the final result value to the specific electronic device. AI server.
  2. delete
  3. delete
  4. delete
  5. In Article 1, The personalized information received from the aforementioned specific electronic device is, One or more weights corresponding to the specific electronic device mentioned above AI server.
  6. In Article 1, The above processor is, Updating the personalization model and transmitting personalization information corresponding to the updated personalization model to the specific electronic device AI server.
  7. In Article 1, The above memory is, A plurality of personalized information corresponding to each of the above plurality of electronic devices is stored, and The above processor is, Receiving input data and identification information corresponding to the specific electronic device from the specific electronic device, obtaining personalized information corresponding to the identification information from the memory, and applying the personalized information corresponding to the identification information to the personalized model. AI server.
  8. In Article 7, The above processor is, Updating the personalization model and storing personalization information corresponding to the updated personalization model in the memory AI server.
  9. In Article 1, The above input data is, video data or audio data AI server.
  10. In Article 1, The above processor is, Receiving input data and environmental information corresponding to the specific electronic device from the specific electronic device, inputting the input data into the common model to obtain an intermediate result value, applying personalized information corresponding to the specific electronic device to the personalized model, and inputting the intermediate result value and the environmental information into the personalized model to which the personalized information is applied to obtain the final result value. AI server.
  11. In Article 10, The above environmental information is, including at least one of the noise magnitude and illuminance AI server.
  12. delete
  13. delete

Description

Artificial Intelligence Server The present invention relates to an artificial intelligence server capable of saving server storage space and providing a personalized recognition service by receiving personalized information from a user who has requested recognition and performing personalized recognition using the received personalized information. Artificial intelligence is a field of computer science and information technology that studies methods to enable computers to perform thinking, learning, and self-development capable of human intelligence, and refers to making computers mimic intelligent human behavior. Furthermore, artificial intelligence does not exist in isolation but is closely related, directly and indirectly, to many other fields of computer science. Particularly in the modern era, there are very active attempts to introduce AI elements into various sectors of information technology and utilize them to solve problems within those fields. Meanwhile, technologies that utilize artificial intelligence to perceive and learn surrounding situations, provide information desired by the user in the desired format, or perform actions or functions desired by the user are being actively researched. And electronic devices that provide these various operations and functions can be referred to as artificial intelligence devices. An artificial intelligence device may include an electronic device used directly by the user and an artificial intelligence server that communicates with the electronic device and provides recognition services to the electronic device. And the artificial intelligence system can be composed of multiple electronic devices and an artificial intelligence server. In an artificial intelligence system, an electronic device receives input data (video, audio, etc.) and transmits it to an artificial intelligence server, and the artificial intelligence server uses the input data to obtain recognition results (such as the presence or absence of obstacles, structure, or objects within the video in the case of video, or the characters corresponding to the audio, the meaning of the audio in the case of audio) and then transmits them to the electronic device. An artificial intelligence server provides recognition services to multiple electronic devices. Generally, the artificial intelligence server inputs input data received from multiple electronic devices into a single deep learning model to obtain recognition results. Meanwhile, the usage environment of multiple electronic devices may vary from device to device. For example, in the case of voice recognition, the first electronic device may be used by a user who speaks a specific regional dialect, the second electronic device may be used by a user who pronounces certain words uniquely, the third electronic device may be used in an environment with a lot of ambient noise, and the fourth electronic device may be used by a user who speaks at a rapid speed. In addition, when providing recognition services using a single deep learning model for multiple electronic devices with different usage environments, there is a disadvantage in that it is difficult to provide personalized recognition services optimized for the usage environment of each electronic device. Meanwhile, in order to optimize for the usage environment of each electronic device, there is a method of operating a deep learning model corresponding to each electronic device. For example, this method involves operating a first deep learning model suitable for a first electronic device used by a user who speaks a dialect of a specific region, a second deep learning model suitable for a second electronic device used by a user who pronounces some words uniquely, a third deep learning model suitable for a third electronic device used in an environment with a lot of ambient noise, and a fourth deep learning model suitable for a fourth electronic device used by a user who speaks at a fast speed, each separately. However, this method has the disadvantage of significantly increasing the storage space required for the server. For example, if the size of a deep learning model for speech recognition is 80MB and 130,000 electronic devices receive speech recognition services from the server, 1TB of storage space is required. Furthermore, when considering the number of product models and special cases for each individual (fast speech speeds, noisy environments, etc.), the problem arises where the actual storage space required increases enormously. FIG. 1a shows an AI device (100) according to one embodiment of the present invention. FIG. 1b shows an AI server (200) according to one embodiment of the present invention. FIG. 2 shows an AI system (1) according to one embodiment of the present invention. FIG. 3 is a block diagram showing the configuration of an electronic device (300) according to one embodiment of the present invention. FIG. 4 is a drawing for explaining an artificial intelligence system acc