WO2018145412A1 - 智能信息处理方法及系统 - Google Patents
智能信息处理方法及系统 Download PDFInfo
- Publication number
- WO2018145412A1 WO2018145412A1 PCT/CN2017/094348 CN2017094348W WO2018145412A1 WO 2018145412 A1 WO2018145412 A1 WO 2018145412A1 CN 2017094348 W CN2017094348 W CN 2017094348W WO 2018145412 A1 WO2018145412 A1 WO 2018145412A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- data
- user
- information processing
- training
- intelligent information
- Prior art date
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
Definitions
- the present invention relates to the field of artificial intelligence technologies, and in particular, to an intelligent information processing method and system.
- the main object of the present invention is to provide an intelligent information processing method for improving the creation efficiency.
- an intelligent information processing method provided by the present invention includes the following steps:
- the training data is used by the cloud knowledge base for machine learning, and receiving prompt data returned from the cloud knowledge base;
- the collecting the original data, the original data including the user's creative content includes:
- the motion information is converted into original data for digitization processing.
- the collecting the original data, the original data including the user's creative content includes:
- the image is converted as raw data for digitization processing.
- the collecting raw data includes the user's creative content:
- the image is converted to the original data for digitization according to an image recognition technique.
- the generating training data for machine learning according to the raw data includes:
- the training data is generated based on the raw data after noise reduction.
- the present invention further provides an intelligent information processing system, where the intelligent information processing system includes: an acquisition module, configured to collect original data, where the original data includes a user's creative content;
- a generating module configured to generate training data for machine learning based on the original data
- a sending module configured to send the training data to a cloud knowledge base, where the training data is used for machine learning by the cloud knowledge base, and receive prompt data returned from the cloud knowledge base;
- the receiving module is configured to receive feedback from the user on the prompt data, and upload the file to the cloud knowledge base.
- the collection module comprises:
- a first obtaining unit configured to acquire motion information of a pen tip for creating the authoring content
- the first conversion unit is configured to convert the motion information into original data for digitization processing.
- the collecting module includes: a second acquiring unit, configured to acquire an image for creating the creative content in real time by using an imaging device;
- a second conversion unit for converting the image as raw data for digitization processing.
- the collection module comprises:
- a third acquiring unit configured to acquire an image of an object that the electronic drawing board is authoring
- a third conversion unit configured to convert the image into the original data for digitization according to an image recognition technology.
- the generating module comprises:
- noise reduction unit configured to perform noise reduction processing on the original data
- a generating unit configured to generate the training data according to the raw data after noise reduction.
- the present invention collects raw data, the original data includes user's authoring content; generates training data for machine learning according to the original data; and sends the training data to a cloud knowledge base, the training data is used for the cloud
- the knowledge base performs machine learning, and receives prompt data returned from the cloud knowledge base; receives feedback from the user on the prompt data, and uploads the data to the cloud knowledge base.
- the user Using the training content of the transformed user as the artificial intelligence training data, the user further confirms the artificial intelligence prompt information, realizes the training of the artificial intelligence, and provides a convenient method for the artificial intelligence training; and at the same time, because the training is in the real situation
- the production is guaranteed to ensure the authenticity of the training; in addition, the user can also obtain the correct prompts during the training process to help the user to create and improve the creation efficiency.
- FIG. 1 is a schematic flowchart of a first embodiment of an intelligent information processing method according to the present invention
- FIG. 2 is a schematic flowchart of a refinement process in a second embodiment of an intelligent information processing method according to the present invention
- FIG. 3 is a schematic flowchart of a refinement process in a third embodiment of an intelligent information processing method according to the present invention.
- FIG. 4 is a schematic diagram of a refinement process in a fourth embodiment of an intelligent information processing method according to the present invention.
- FIG. 5 is a schematic flowchart of a refinement process in a fifth embodiment of an intelligent information processing method according to the present invention.
- FIG. 6 is a schematic diagram of functional modules of a first embodiment of an intelligent information processing system according to the present invention.
- FIG. 7 is a schematic diagram of a refinement function module of an acquisition module in a second embodiment of the intelligent information processing system of the present invention.
- FIG. 8 is a schematic diagram of a refinement function module of an acquisition module in a third embodiment of the intelligent information processing system of the present invention.
- FIG. 9 is a schematic diagram of a refinement function module of an acquisition module in a fourth embodiment of the intelligent information processing system of the present invention.
- FIG. 10 is a schematic diagram of a refinement function module of a generation module in a fifth embodiment of the intelligent information processing system of the present invention.
- the intelligent information processing method includes:
- Step S10 collecting original data, where the original data includes the created content of the user
- the user's creative content is acquired through the sensor, for example, corresponding to the written content, the content of the painting, the content expressed by the limb, the content of the melody, and the like.
- Step S20 generating training data for machine learning according to the original data
- the data is further processed to achieve training data that can be used for artificial intelligence training.
- Step S30 Send the training data to a cloud knowledge base, where the training data is used for machine learning by the cloud knowledge base, and receive prompt data returned from the cloud knowledge base;
- the training data acquisition and machine learning training are run on different machines, and the machine learning and acquisition training data use resources do not interfere with each other to avoid conflicts; at the same time, the cloud knowledge base has been
- Step S40 Receive feedback of the prompt data by the user, and upload the file to the cloud knowledge base.
- the artificial intelligence of the cloud knowledge base is further trained by feedback of the user's prompt information.
- the cloud knowledge base is provided with artificial intelligence to help people improve the drawing ability.
- the user draws the content drawn by the user, generates artificial intelligence training data according to the content, and uploads.
- the cloud knowledge base analyzes the content of the painting, it outputs the artificial intelligence's own creation result.
- the cloud knowledge base returns a rabbit picture to the user, if the user Originally wanted to draw a rabbit, then the artificial intelligence sent back to the cloud knowledge base, the output is correct; if the user originally wanted to draw a horse, then the artificial intelligence fed back to the cloud knowledge base, the output result is wrong, training artificial intelligence acquisition The feature vector of the rabbit in the painting.
- the user can further confirm the artificial intelligence prompt information, realize the training of artificial intelligence, and provide a convenient method for artificial intelligence training; at the same time, because the training is in the real situation
- the production is guaranteed to ensure the authenticity of the training; in addition, the user can also obtain the correct prompts during the training process to help the user to create and improve the creation efficiency.
- the step S10 includes:
- Step S11 acquiring motion information of a pen tip for creating the authoring content
- step S12 the motion information is converted into original data for digitization processing.
- a gyroscope is installed in the nib, and the writing content of the user is analyzed by continuously acquiring information such as the moving direction, moving speed, and moving time of the nib, and the writing content of the user is converted into an electronic document. Data, and further converting the electronic document data to training data for training artificial intelligence. It should be noted that in some embodiments a pressure sensor is used on the user writing container to obtain the speed of movement of the nib.
- the pen tip creation content is a stroke
- the pen tip motion information is not fully input by the user
- the pen tip motion information is converted into original data for digitization processing, and uploaded to the cloud knowledge base, cloud knowledge.
- the library uses artificial intelligence to determine the text that the user wants to write, and returns it to the user. If the user confirms, the artificial intelligence is correctly judged, and the artificial intelligence is trained, which further enhances the user experience.
- the pen tip creation content is a picture, and after each preset time, automatically uploading data according to the pen tip motion information to the cloud knowledge base, and the cloud knowledge base uses artificial intelligence to determine that the user wants to draw And returning the content to determine the content that the user wants to draw to the client for reference by the user. If the user confirms, the artificial intelligence is correctly judged, and the artificial intelligence is trained, thereby further improving the user experience.
- the step S10 includes:
- Step S13 acquiring an image for creating the created content in real time by using an imaging device
- step S14 the image is converted as original data for digitization processing.
- the user action is obtained to predict the user behavior. For example, after the user picks up the baton and makes the specific gesture data, it is transmitted to the cloud knowledge base. After analysis, the user wants to practice the result of the music command, and further observes whether the user performs the command practice by playing the preset music. As feedback, if the user follows the music practice command, the machine judges correctly; if the baton is put down, the machine judges the error and then trains the artificial intelligence according to the above feedback.
- the camera continuously acquires the image of the user, analyzes the image, obtains motion data of the user, converts the motion data into training data for training artificial intelligence, accelerates training of artificial intelligence, and simultaneously uses artificial intelligence. Predict the user's actions and provide users with corresponding services in advance to enhance the user experience.
- the user-drawn picture acquired by the camera device in real time converts the picture drawn by the user into original data and uploads the picture to the cloud knowledge base, and the cloud knowledge base is analyzed to determine the user wants to draw.
- Contenting, and generating a prompting picture to the client if the content drawn by the user in the next step is consistent with the prompting picture, the machine judges correctly; if the prompting picture is inconsistent with the content drawn by the user in the next step, The machine judges the error and trains the artificial intelligence according to the above feedback to further enhance the user experience.
- the content written by the user acquired by the camera in real time converts the content written by the user into original data and uploads the picture to the cloud knowledge base, and the cloud knowledge base is analyzed to determine that the user wants Writing the content and generating the prompting text to the client, if the text written by the user in the next step is consistent with the prompting text, the machine judges correctly; if the prompting image is inconsistent with the content drawn by the user in the next step The machine judges the error and trains the artificial intelligence according to the above feedback to further enhance the user experience.
- the step S10 includes:
- Step S15 acquiring an image of an object that the electronic drawing board is creating
- Step S16 converting the image into the original data for digitization processing according to an image recognition technology.
- the user draws on the electronic drawing board, obtains the user's graphic on the electronic drawing board, and sends the artificial intelligence to the cloud knowledge base in real time according to the direction of the newly generated line, and allows the artificial intelligence to make an output.
- the graphics and colors on the electronic drawing board generate the content predicted in the next step, and judge the output of the artificial intelligence through the user to speed up the training of artificial intelligence; at the same time, the output prompt of artificial intelligence helps the user to improve Drawing efficiency.
- the step S20 includes:
- Step S21 performing noise reduction processing on the original data
- the original data is analyzed to remove unnecessary features, such as monitoring the nib and acquiring the written content, only extract part of the text as the training object; , automatically removes the background elements of the picture.
- Step S22 generating the training data according to the raw data after noise reduction.
- the background elements in the drawing content are automatically removed, only the center of the drawing is retained, and the cropped image is used as the training data.
- the intelligent information processing system includes:
- the collecting module 10 is configured to collect original data, where the original data includes the created content of the user;
- the user's creative content is acquired through the sensor, for example, corresponding to the written content, the content of the painting, the content expressed by the limb, the content of the melody, and the like.
- a generating module 20 configured to generate training data for machine learning according to the original data
- the data is further processed to achieve training data that can be used for artificial intelligence training.
- the sending module 30 is configured to send the training data to a cloud knowledge base, where the training data is used for machine learning by the cloud knowledge base, and receive prompt data returned from the cloud knowledge base;
- the training data acquisition and machine learning training are run on different machines, and the machine learning and acquisition training data use resources do not interfere with each other to avoid conflicts; at the same time, the cloud knowledge base has been
- the receiving module 40 is configured to receive feedback of the prompt data by the user, and upload the file to the cloud knowledge base.
- the artificial intelligence of the cloud knowledge base is further trained by feedback of the user's prompt information.
- the cloud knowledge base is provided with artificial intelligence to help people improve the drawing ability.
- the collecting module 10 acquires the content drawn by the user through the camera, and the generating module 20 generates the artificial according to the content.
- Intelligent training data after the sending module 30 uploads to the cloud knowledge base to analyze the content of the painting, the receiving module 40 outputs the artificial intelligence own creation result, and more, when the user draws the two animal ears, the cloud knowledge The library returns a picture of the rabbit to the user.
- the artificial intelligence fed back to the cloud knowledge base will output the correct result; if the user originally wanted to draw the horse, the artificial intelligence fed back to the cloud knowledge base The output is wrong, and the artificial intelligence is trained to obtain the feature vector of the rabbit in the painting.
- the user can further confirm the artificial intelligence prompt information, realize the training of artificial intelligence, and provide a convenient method for artificial intelligence training; at the same time, because the training is in the real situation The production is guaranteed to ensure the authenticity of the training; in addition, the user can also obtain the correct prompts during the training process to help the user to create and improve the creation efficiency.
- the collection module 10 includes:
- a first obtaining unit 11 configured to acquire motion information of a pen tip for creating the authoring content
- the first converting unit 12 is configured to convert the motion information into original data for digitization processing.
- the gyroscope is installed in the pen tip, and the first acquiring unit 11 analyzes the writing content of the user by continuously acquiring the moving direction, the moving speed, the moving time and the like of the pen tip, and the first converting unit 12 Converting the written content of the user into electronic document data, and further converting the electronic document data into training data for training artificial intelligence.
- the pressure sensor is used on the user writing container to obtain the moving speed of the pen tip.
- the nib creation content is a stroke
- the first acquisition unit 11 converts the nib motion information into original data for digitization processing, the nib motion information that the user has not fully input.
- the cloud knowledge base uses artificial intelligence to judge the text that the user wants to write, and returns it to the user. If the user confirms, the artificial intelligence is correctly judged, and the artificial intelligence is trained, which further improves the user experience.
- the pen tip creation content is a picture
- the first acquiring unit 11 acquires motion information of the pen tip within a preset time by the user
- the first conversion unit 12 converts the
- the pen tip motion information is the original data used for digitization processing, and is uploaded to the cloud knowledge base.
- the cloud knowledge base uses artificial intelligence to determine the content that the user wants to draw, and returns the content that is determined by the user to be drawn to the client for the user. For reference, if the user confirms, the artificial intelligence is correctly judged, and the artificial intelligence is trained, which further improves the user experience.
- the collection module 10 includes:
- a second acquiring unit 13 configured to acquire an image for creating the creation in real time by using an imaging device
- the second converting unit 14 is configured to convert the image as original data for digitization processing.
- the user action is obtained to predict the user behavior. For example, after the user picks up the baton and makes the specific gesture data, it is transmitted to the cloud knowledge base. After analysis, the user wants to practice the result of the music command, and further observes whether the user performs the command practice by playing the preset music. As feedback, if the user follows the music practice command, the machine judges correctly; if the baton is put down, the machine judges the error and then trains the artificial intelligence according to the above feedback.
- the second acquiring unit 13 continuously acquires the image of the user through the camera, and the second converting unit 14 analyzes the image to obtain motion data of the user, and converts the motion data into training data for training artificial intelligence, which speeds up. Artificial intelligence training; at the same time, by using artificial intelligence to predict the user's actions, and providing users with corresponding services in advance, the user experience is improved.
- the second obtaining unit 13 uses the user-drawn picture acquired by the camera device in real time, and the second converting unit 14 converts the picture drawn by the user into original data and uploads the picture to the cloud knowledge base, in the cloud.
- the knowledge base is analyzed to obtain the content that the user wants to draw, and generates a prompt image to be sent to the client. If the content that the user draws next matches the prompt image, the machine judges correctly; if the prompt image and If the content drawn by the user in the next step is inconsistent, the machine judges the error, and the artificial intelligence is trained according to the feedback to further improve the user experience.
- the second obtaining unit 13 uses the content written by the user acquired by the camera in real time
- the second converting unit 14 converts the content written by the user into original data and uploads the drawing to the cloud knowledge base.
- the cloud knowledge base analyzes the content that the user wants to write, and generates prompt text to be sent to the client. If the text written by the user and the prompt text match, the machine judges correctly; if the prompt is If the picture is inconsistent with the content drawn by the user in the next step, the machine judges the error, and the artificial intelligence is trained according to the above feedback to further enhance the user experience.
- the collection module 10 includes:
- the third obtaining unit 15 acquires an image of the object that the electronic drawing board is authoring
- the third converting unit 16 converts the image into the original data for digitization processing according to an image recognition technique.
- the third acquisition unit 15 acquires the graphic of the user on the electronic drawing board, and the third conversion unit 16 generates the original data according to the newly generated line, and sends the data to the cloud knowledge base in real time.
- Artificial intelligence and let the artificial intelligence make the output, directly through the graphics and color on the electronic drawing board, generate the content of the next prediction, and judge the output of the artificial intelligence through the user, which accelerates the training of artificial intelligence;
- the artificial intelligence output prompts, and helps users improve the drawing efficiency.
- the generating module 20 includes:
- noise reduction unit 21 configured to perform noise reduction processing on the original data
- the original data is analyzed to remove unnecessary features, such as monitoring the nib and acquiring the written content, only extract part of the text as the training object; , automatically removes the background elements of the picture.
- the generating unit 22 is configured to generate the training data according to the raw data after noise reduction.
- the noise reduction unit 21 automatically removes the background element in the drawing content, leaving only the drawing center area, and the generating unit 22 uses the cropped picture as the training data.
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Software Systems (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Medical Informatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Artificial Intelligence (AREA)
- Image Analysis (AREA)
Abstract
本发明公开了一种智能信息处理方法。所述智能信息处理方法包括以下步骤：采集原始数据；根据原始数据生成用于机器学习的训练数据；发送训练数据至云端知识库，训练数据用于云端知识库进行机器学习，并接收从云端知识库返回的提示数据；接收用户对提示数据的反馈，并上传至云端知识库。本发明还公开了一种智能信息处理系统。本发明通过转化用户的创作内容为人工智能的训练数据，并让用户对提示信息进行进一步确认，实现对人工智能的训练，为人工智能的训练提供了方便的方法；同时因为训练是在真实情况下产生，保证了训练的真实性；此外，训练过程中用户也可以随时获取正确的提示，帮助用户进行创作，提高创作效率。
Description
技术领域
本发明涉及人工智能技术领域，尤其涉及智能信息处理方法及系统。
背景技术
在人们进行例如写作、绘画等创作时，根据自己的思路进行创作，但是由于各种因素的影响，创作思路出现中断时，需要花时间思考，例如在写作时不能不知道使用哪些词语，或者在绘画时，不知道使用什么样的线条或者颜色，这往往都需要花时间进行思考，影响创作效率。
发明内容
本发明的主要目的在于提供一种智能信息处理方法，旨提高创作效率。
为实现上述目的，本发明提供的一种智能信息处理方法包括以下步骤：
采集原始数据，所述原始数据包括用户的创作内容；
根据所述原始数据生成用于机器学习的训练数据；
发送所述训练数据至云端知识库，所述训练数据用于所述云端知识库进行机器学习，并接收从所述云端知识库返回的提示数据；
接收用户对所述提示数据的反馈，并上传至所述云端知识库。
优选地，所述采集原始数据，所述原始数据包括用户的创作内容包括：
获取用于创作所述创作内容的笔尖的运动信息；
转换所述运动信息为用于数字化处理的原始数据。
优选地，所述采集原始数据，所述原始数据包括用户的创作内容包括：
通过摄像装置实时获取用于创作所述创作内容的影像；
转换所述影像作为用于数字化处理的原始数据。
优选地，所述采集原始数据，所述原始数据包括用户的创作内容：
获取电子画板正在创作对象的图像；
根据图像识别技术转换所述图像为用于数字化处理的所述原始数据。
优选地，所述根据所述原始数据生成用于机器学习的训练数据包括；
对所述原始数据进行降噪处理；
根据降噪后的所述原始数据生成所述训练数据。
此外，为实现上述目的，本发明还提供一种智能信息处理系统，所述智能信息处理系统包括：采集模块，用于采集原始数据，所述原始数据包括用户的创作内容；
生成模块，用于根据所述原始数据生成用于机器学习的训练数据；
发送模块，用于发送所述训练数据至云端知识库，所述训练数据用于所述云端知识库进行机器学习，并接收从所述云端知识库返回的提示数据；
接收模块，用于接收用户对所述提示数据的反馈，并上传至所述云端知识库。
优选地，所述采集模块包括：
第一获取单元，用于获取用于创作所述创作内容的笔尖的运动信息；
第一转换单元，用于转换所述运动信息为用于数字化处理的原始数据。
优选地，所述采集模块包括：第二获取单元，用于通过摄像装置实时获取用于创作所述创作内容的影像；
第二转换单元，用于转换所述影像作为用于数字化处理的原始数据。
优选地，所述采集模块包括：
第三获取单元，用于获取电子画板正在创作对象的图像；
第三转换单元，用于根据图像识别技术转换所述图像为用于数字化处理的所述原始数据。
优选地，所述生成模块包括：
降噪单元，用于对所述原始数据进行降噪处理；
生成单元，用于根据降噪后的所述原始数据生成所述训练数据。
本发明通过采集原始数据，所述原始数据包括用户的创作内容；根据所述原始数据生成用于机器学习的训练数据；发送所述训练数据至云端知识库，所述训练数据用于所述云端知识库进行机器学习，并接收从所述云端知识库返回的提示数据；接收用户对所述提示数据的反馈，并上传至所述云端知识库。使用转化用户的创作内容为人工智能的训练数据，还使用户对人工智提示信息的进一步确认，实现对人工智能的训练，为人工智能的训练提供了方便的方法；同时因为训练是在真实情况下产生，保证了训练的真实性；此外训练过程中用户也可以随时获取正确的提示，帮助用户进行创作，提高创作效率。
附图说明
图1为本发明智能信息处理方法第一实施例的流程示意图；
图2为本发明智能信息处理方法第二实施例中的细化流程示意图；
图3为本发明智能信息处理方法第三实施例中的细化流程示意图；
图4为本发明智能信息处理方法第四实施例中细化流程示意图；
图5为本发明智能信息处理方法第五实施例中的细化流程示意图；
图6为本发明智能信息处理系统第一实施例的功能模块示意图；
图7为本发明智能信息处理系统第二实施例中采集模块的细化功能模块示意图；
图8为本发明智能信息处理系统第三实施例中采集模块的细化功能模块示意图；
图9为本发明智能信息处理系统第四实施例中采集模块的细化功能模块示意图；
图10为本发明智能信息处理系统第五实施例中生成模块的细化功能模块示意图。
本发明目的的实现、功能特点及优点将结合实施例，参照附图做进一步说明。
具体实施方式
应当理解，此处所描述的具体实施例仅仅用以解释本发明，并不用于限定本发明。
本发明提供一种智能信息处理方法，请参照图1，在本发明智能信息处理方法第一实施例中，该智能信息处理方法包括：
步骤S10，采集原始数据，所述原始数据包括用户的创作内容；
在用户在进行创作时，通过传感器获取用户的创作内容，例如：对应为书写的内容，绘画的内容，肢体表达出的内容，吹奏旋律的内容等。
步骤S20，根据所述原始数据生成用于机器学习的训练数据；
通过传感器把用户创作的内容转化为原始数据后，进一步对数据进行处理，使其达到可用于进行人工智能训练的训练数据。
步骤S30，发送所述训练数据至云端知识库，所述训练数据用于所述云端知识库进行机器学习，并接收从所述云端知识库返回的提示数据；
通过把云端知识库设置在远端服务器，使得训练数据的获得和机器学习的训练在不同的机器上运行，机器学习与获取训练数据使用资源互不干扰，避免冲突；同时在云端知识库根据已经获得的人工智能对原始数据进行判断，并生成提示数据，例如，当训练数据为手写文本“1+1”时，提示数据为“=2”。
步骤S40，接收用户对所述提示数据的反馈，并上传至所述云端知识库。
通过根据用户对提示信息的反馈，进一步训练所述云端知识库的人工智能。具体在本实施例中，所述云端知识库设有帮助人提高绘画能力的人工智能，用户端进行绘画时，通过相机获取用户所画的内容，根据所述内容生成人工智能训练数据，并上传至云端知识库对所述绘画的内容进行分析后，输出人工智能自己的创作结果，更具地，当获取用户绘制完成两个动物耳朵后，云端知识库返回一个兔子的图画给用户，如果用户本来想画兔子，则反馈给云端知识库的人工智能，所输出的结果正确；如果用户本来想画的是马，则反馈给云端知识库的人工智能，所输出的结果错误，训练人工智能获取绘画中兔子的特征向量。通过转化用户的创作内容为人工智能的训练数据，还让用户对人工智提示信息的进一步确认，实现对人工智能的训练，为人工智能的训练提供了方便的方法；同时因为训练是在真实情况下产生，保证了训练的真实性；此外训练过程中用户也可以随时获取正确的提示，帮助用户进行创作，提高创作效率。
请参照图2，基于本发明智能信息处理方法第一实施例，在本发明智能信息处理方法第二实施例中，所述步骤S10包括：
步骤S11，获取用于创作所述创作内容的笔尖的运动信息；
步骤S12，转换所述运动信息为用于数字化处理的原始数据。
在本实施例中，所述笔尖内安装有陀螺仪，通过持续获取所述笔尖的移动方向，移动速度，移动时间等信息，分析出用户的书写内容，转换所述用户的书写内容为电子文档数据，并进一步转换所述电子文档数据为训练人工智能的训练数据。应当说明的是，在有些实施例中在用户写作容器上使用压力传感器获取所述笔尖的移动速度。
通过把笔尖的运动信息转换为训练人工智能的原始数据，因为通过笔尖书写是一种普遍行为，所以数据量大，使得在训练有关文本方面的人工智能时，能获得足够多的数据，加快了人工智能的训练；同时通过为用户提供人工智能的提示，也提升了用户书写时的体验。
在有些实施例中，所述笔尖创作内容为为笔画，在用户还未完全输入的笔尖运动信息，转化所述笔尖运动信息为用于数字化处理的原始数据，并上传至云端知识库，云端知识库使用人工智能判断出用户所要书写的文字，并返回给用户，如果用户确认则所述人工智能判断正确，练了人工智能，也进一步提升了用户体验。
在另一些实施例中，所述笔尖创作内容为为图画，在每个预设时间后，自动上传根据所述笔尖运动信息产生数据至云端知识库，云端知识库使用人工智能判断出用户所要绘制的内容，并返回所述判断出用户所要绘制的内容至客户端供用户参考，如果用户确认则所述人工智能判断正确，练了人工智能，也进一步提升了用户体验。
请参照图3，基于本发明智能信息处理方法第一实施例，在本发明智能信息处理方法第三实施例中，所述步骤S10包括：
步骤S13，通过摄像装置实时获取用于创作所述创作内容的影像；
步骤S14，转换所述影像作为用于数字化处理的原始数据。
更具体地，通过监测用户影像，获取用户动作预判用户行为。例如：得到用户在拿起指挥棒并作出特定起手势的数据后传至云端知识库，经过分析得出用户想要练习乐曲指挥的结果，并通过播放预设的音乐进一步观察用户是否进行指挥练习作为反馈，如果用户跟随乐曲练习指挥，则机器判断正确；如果放下指挥棒，则机器判断错误，进而根据上述反馈训练人工智能。
在本实施例中，通过摄像机持续获取用户的影像，分析所述影像，得到用户的动作数据，转换所述动作数据为训练人工智能的训练数据，加快了人工智能的训练；同时通过使用人工智能预判用户的动作，并为用户提前提供相应服务，提升了用户体验。
在有些实施例中，所述摄像装置实时获取的用户绘制的图画，转化所述用户绘制的图画为原始数据并上传所述图画至云端知识库，云端知识库经过分析得出用户想要绘制的内容，并生成提示性图片发给客户端，如果用户下一步绘制的内容和所述提示性图片相符合则机器判断正确；如果所述提示性图片和所述用户下一步绘制的内容不一致，则机器判断错误，根据上述反馈训练人工智能，进一步提升用户体验。
而在另一些实施例中，所述摄像装置实时获取的用户书写的内容，转化所述用户书写的内容为原始数据并上传所述图画至云端知识库，云端知识库经过分析得出用户想要书写的内容，并生成提示性文字发给客户端，如果用户下一步书写的文字和所述提示性文字相符合则机器判断正确；如果所述提示性图片和所述用户下一步绘制的内容不一致，则机器判断错误，根据上述反馈训练人工智能，进一步提升用户体验。
请参照图4，基于本发明智能信息处理方法第一实施例，在本发明智能信息处理方法第四实施例中，所述步骤S10包括：
步骤S15，获取电子画板正在创作对象的图像；
步骤S16，根据图像识别技术转换所述图像为用于数字化处理的所述原始数据。
在本实施例中，用户在电子画板上的作图，获取用户在所述电子画板上的图形，根据新生成线条的走向，实时发送给云端知识库的人工智能，并让人工智能做出输出，直接通过电子画板上的图形和颜色，生成下一步预测的内容，仔经过用户对所述人工智能的输出结果进行判断，加快了人工智能的训练；同时人工智能的输出提示，又帮助用户提高了作图效率。
请参照图5，基于本发明智能信息处理方法第一实施例，在本发明智能信息处理方法第五实施例中，所述步骤S20包括：
步骤S21，对所述原始数据进行降噪处理；
在对人工智能进行训练时，为了减少计算量，对原始数据进行分析，去除不必要的特征，例如在对笔尖进行监测，并获取书写内容时，只提取部分的文本作为训练对象；在绘画时，自动去除画面的背景元素。
步骤S22，根据降噪后的所述原始数据生成所述训练数据。
在本实施例中，在获取用户的绘画内容后，自动去除绘画内容中的背景元素，只保留绘画中心区域，并使用剪裁过后的图片作为训练数据。通过对原始数据进行降噪，进一步提高了训练的质量，减少机器学习的时间；同时也提高了机器学习的准确度。
本发明提供一种智能信息处理系统，请参照图6，在本发明智能信息处理方法第一实施例中，该智能信息处理系统包括：
采集模块10，用于采集原始数据，所述原始数据包括用户的创作内容；
在用户在进行创作时，通过传感器获取用户的创作内容，例如：对应为书写的内容，绘画的内容，肢体表达出的内容，吹奏旋律的内容等。
生成模块20，用于根据所述原始数据生成用于机器学习的训练数据；
通过传感器把用户创作的内容转化为原始数据后，进一步对数据进行处理，使其达到可用于进行人工智能训练的训练数据。
发送模块30，用于发送所述训练数据至云端知识库，所述训练数据用于所述云端知识库进行机器学习，并接收从所述云端知识库返回的提示数据；
通过把云端知识库设置在远端服务器，使得训练数据的获得和机器学习的训练在不同的机器上运行，机器学习与获取训练数据使用资源互不干扰，避免冲突；同时在云端知识库根据已经获得的人工智能对原始数据进行判断，并生成提示数据，例如，当训练数据为手写文本“1+1”时，提示数据为“=2”。
接收模块40，用于接收用户对所述提示数据的反馈，并上传至所述云端知识库。
通过根据用户对提示信息的反馈，进一步训练所述云端知识库的人工智能。具体在本实施例中，所述云端知识库设有帮助人提高绘画能力的人工智能，用户端进行绘画时，采集模块10通过相机获取用户所画的内容，生成模块20根据所述内容生成人工智能训练数据，发送模块30上传至云端知识库对所述绘画的内容进行分析后，接收模块40输出人工智能自己的创作结果，更具地，当获取用户绘制完成两个动物耳朵后，云端知识库返回一个兔子的图画给用户，如果用户本来想画兔子，则反馈给云端知识库的人工智能，所输出的结果正确；如果用户本来想画的是马，则反馈给云端知识库的人工智能，所输出的结果错误，训练人工智能获取绘画中兔子的特征向量。通过转化用户的创作内容为人工智能的训练数据，还让用户对人工智提示信息的进一步确认，实现对人工智能的训练，为人工智能的训练提供了方便的方法；同时因为训练是在真实情况下产生，保证了训练的真实性；此外训练过程中用户也可以随时获取正确的提示，帮助用户进行创作，提高创作效率。
请参照图7，基于本发明智能信息处理系统第一实施例，在本发明智能信息处理方法系统第二实施例中，所述采集模块10包括：
第一获取单元11，用于获取用于创作所述创作内容的笔尖的运动信息；
第一转换单元12，用于转换所述运动信息为用于数字化处理的原始数据。
在本实施例中，所述笔尖内安装有陀螺仪，第一获取单元11通过持续获取所述笔尖的移动方向，移动速度，移动时间等信息，分析出用户的书写内容，第一转换单元12转换所述用户的书写内容为电子文档数据，并进一步转换所述电子文档数据为训练人工智能的训练数据。
通过把笔尖的运动信息转换为训练人工智能的原始数据，因为通过笔尖书写是一种普遍行为，所以数据量大，使得在训练有关文本方面的人工智能时，能获得足够多的数据，加快了人工智能的训练；同时通过为用户提供人工智能的提示，也提升了用户书写时的体验，应当说明的是，在有些实施例中在用户写作容器上使用压力传感器获取所述笔尖的移动速度。
在有些实施例中，所述笔尖创作内容为为笔画，第一获取单元11在用户还未完全输入的笔尖运动信息，第一转换单元12转化所述笔尖运动信息为用于数字化处理的原始数据，并上传至云端知识库，云端知识库使用人工智能判断出用户所要书写的文字，并返回给用户，如果用户确认则所述人工智能判断正确，练了人工智能，也进一步提升了用户体验。
在另一些实施例中，所述笔尖创作内容为为图画，在每个预设时间后，第一获取单元11在用户获取预设时间内所述笔尖的运动信息，第一转换单元12转化所述笔尖运动信息为用于数字化处理的原始数据，并上传至云端知识库，云端知识库使用人工智能判断出用户所要绘制的内容，并返回所述判断出用户所要绘制的内容至客户端供用户参考，如果用户确认则所述人工智能判断正确，练了人工智能，也进一步提升了用户体验。
请参照图8，基于本发明智能信息处理系统第一实施例，在本发明智能信息处理系统第三实施例中，所述采集模块10包括：
第二获取单元13，用于通过摄像装置实时获取用于创作所述创作的影像；
第二转换单元14，用于转换所述影像作为用于数字化处理的原始数据。
更具体地，通过监测用户影像，获取用户动作预判用户行为。例如：得到用户在拿起指挥棒并作出特定起手势的数据后传至云端知识库，经过分析得出用户想要练习乐曲指挥的结果，并通过播放预设的音乐进一步观察用户是否进行指挥练习作为反馈，如果用户跟随乐曲练习指挥，则机器判断正确；如果放下指挥棒，则机器判断错误，进而根据上述反馈训练人工智能。
在本实施例中，第二获取单元13通过摄像机持续获取用户的影像，第二转换单元14分析所述影像，得到用户的动作数据，转换所述动作数据为训练人工智能的训练数据，加快了人工智能的训练；同时通过使用人工智能预判用户的动作，并为用户提前提供相应服务，提升了用户体验。
在有些实施例中，第二获取单元13使用所述摄像装置实时获取的用户绘制的图画，第二转换单元14转化所述用户绘制的图画为原始数据并上传所述图画至云端知识库，云端知识库经过分析得出用户想要绘制的内容，并生成提示性图片发给客户端，如果用户下一步绘制的内容和所述提示性图片相符合则机器判断正确；如果所述提示性图片和所述用户下一步绘制的内容不一致，则机器判断错误，根据上述反馈训练人工智能，进一步提升用户体验。
而在另一些实施例中，第二获取单元13使用所述摄像装置实时获取的用户书写的内容，第二转换单元14转化所述用户书写的内容为原始数据并上传所述图画至云端知识库，云端知识库经过分析得出用户想要书写的内容，并生成提示性文字发给客户端，如果用户下一步书写的文字和所述提示性文字相符合则机器判断正确；如果所述提示性图片和所述用户下一步绘制的内容不一致，则机器判断错误，根据上述反馈训练人工智能，进一步提升用户体验。
请参照图9，基于本发明智能信息处理系统第一实施例，在本发明智能信息处理系统第四实施例中，所述采集模块10包括：
第三获取单元15，获取电子画板正在创作对象的图像；
第三转换单元16，根据图像识别技术转换所述图像为用于数字化处理的所述原始数据。
在本实施例中，用户在电子画板上的作图，第三获取单元15获取用户在所述电子画板上的图形，第三转换单元16根据新生成线条生成原始数据，实时发送给云端知识库的人工智能，并让人工智能做出输出，直接通过电子画板上的图形和颜色，生成下一步预测的内容，仔经过用户对所述人工智能的输出结果进行判断，加快了人工智能的训练；同时人工智能的输出提示，又帮助用户提高了作图效率。
请参照图10，基于本发明智能信息处理系统第一实施例，在本发明智能信息处理系统第五实施例中，所述生成模块20包括：
降噪单元21，用于对所述原始数据进行降噪处理；
在对人工智能进行训练时，为了减少计算量，对原始数据进行分析，去除不必要的特征，例如在对笔尖进行监测，并获取书写内容时，只提取部分的文本作为训练对象；在绘画时，自动去除画面的背景元素。
生成单元22，用于根据降噪后的所述原始数据生成所述训练数据。
在本实施例中，在获取用户的绘画内容后，降噪单元21自动去除绘画内容中的背景元素，只保留绘画中心区域，生成单元22使用剪裁过后的图片作为训练数据。通过对原始数据进行降噪，进一步提高了训练的质量，减少机器学习的时间；同时也提高了机器学习的准确度。
以上仅为本发明的优选实施例，并非因此限制本发明的专利范围，凡是利用本发明说明书及附图内容所作的等效结构或等效流程变换，或直接或间接运用在其他相关的技术领域，均同理包括在本发明的专利保护范围内。
Claims (16)
- 一种智能信息处理方法，其特征在于，所述智能信息处理方法包括以下步骤：采集原始数据，所述原始数据包括用户的创作内容；根据所述原始数据生成用于机器学习的训练数据；发送所述训练数据至云端知识库，所述训练数据用于所述云端知识库进行机器学习，并接收从所述云端知识库返回的提示数据；接收用户对所述提示数据的反馈，并上传至所述云端知识库。
- 如权利要求1所述的智能信息处理方法，其特征在于，所述根据所述原始数据生成用于机器学习的训练数据包括：对所述原始数据进行降噪处理；根据降噪后的所述原始数据生成所述训练数据。
- 如权利要求1所述的智能信息处理方法，其特征在于，所述采集原始数据，所述原始数据包括用户的创作内容包括：获取用于创作所述创作内容的笔尖的运动信息；转换所述运动信息为用于数字化处理的原始数据。
- 如权利要求3所述的智能信息处理方法，其特征在于，所述根据所述原始数据生成用于机器学习的训练数据包括：对所述原始数据进行降噪处理；根据降噪后的所述原始数据生成所述训练数据。
- 如权利要求1所述的智能信息处理方法，其特征在于，所述采集原始数据，所述原始数据包括用户的创作内容包括：通过摄像装置实时获取用于创作所述创作内容的影像；换所述影像为用于数字化处理的原始数据。
- 如权利要求5所述的智能信息处理方法，其特征在于，所述根据所述原始数据生成用于机器学习的训练数据包括：对所述原始数据进行降噪处理；根据降噪后的所述原始数据生成所述训练数据。
- 如权利要求1所述的智能信息处理方法，其特征在于，所述采集原始数据，所述原始数据包括用户的创作内容：获取电子画板正在创作对象的图像；根据图像识别技术转换所述图像为用于数字化处理的所述原始数据。
- 如权利要求7所述的智能信息处理方法，其特征在于，所述根据所述原始数据生成用于机器学习的训练数据包括：对所述原始数据进行降噪处理；根据降噪后的所述原始数据生成所述训练数据。
- 一种智能信息处理系统，其特征在于，所述智能信息处理系统包括：采集模块，用于采集原始数据，所述原始数据包括用户的创作内容；生成模块，用于根据所述原始数据生成用于机器学习的训练数据；发送模块，用于发送所述训练数据至云端知识库，所述训练数据用于所述云端知识库进行机器学习，并接收从所述云端知识库返回的提示数据；接收模块，用于接收用户对所述提示数据的反馈，并上传至所述云端知识库。
- 如权利要求9所述的智能信息处理系统，其特征在于，所述生成模块包括：降噪单元，用于对所述原始数据进行降噪处理；生成单元，用于根据降噪后的所述原始数据生成所述训练数据。
- 如权利要求9所述的智能信息处理系统，其特征在于，所述采集模块包括：第一获取单元，用于获取用于创作所述创作内容的笔尖的运动信息；第一转换单元，用于转换所述运动信息为用于数字化处理的原始数据。
- 如权利要求11所述的智能信息处理系统，其特征在于，所述生成模块包括：降噪单元，用于对所述原始数据进行降噪处理；生成单元，用于根据降噪后的所述原始数据生成所述训练数据。
- 如权利要求9所述的智能信息处理系统，其特征在于，所述采集模块包括：第二获取单元，用于通过摄像装置实时获取用于创作所述创作内容的影像；第二转换单元，用于转换所述影像作为用于数字化处理的原始数据。
- 如权利要求13所述的智能信息处理系统，其特征在于，所述生成模块包括：降噪单元，用于对所述原始数据进行降噪处理；生成单元，用于根据降噪后的所述原始数据生成所述训练数据。
- 如权利要求9所述的智能信息处理系统，其特征在于，所述采集模块包括：第三获取单元，用于获取电子画板正在创作对象的图像；第三转换单元，用于根据图像识别技术转换所述图像为用于数字化处理的所述原始数据。
- 如权利要求15所述的智能信息处理系统，其特征在于，所述生成模块包括：降噪单元，用于对所述原始数据进行降噪处理；生成单元，用于根据降噪后的所述原始数据生成所述训练数据。
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710068552.X | 2017-02-08 | ||
| CN201710068552.XA CN106886820A (zh) | 2017-02-08 | 2017-02-08 | 智能信息处理方法及系统 |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| WO2018145412A1 true WO2018145412A1 (zh) | 2018-08-16 |
Family
ID=59178968
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| PCT/CN2017/094348 WO2018145412A1 (zh) | 2017-02-08 | 2017-07-25 | 智能信息处理方法及系统 |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN106886820A (zh) |
| WO (1) | WO2018145412A1 (zh) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106886820A (zh) * | 2017-02-08 | 2017-06-23 | 深圳市科迈爱康科技有限公司 | 智能信息处理方法及系统 |
| US11321611B2 (en) | 2018-10-03 | 2022-05-03 | International Business Machines Corporation | Deployment verification of authenticity of machine learning results |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5974412A (en) * | 1997-09-24 | 1999-10-26 | Sapient Health Network | Intelligent query system for automatically indexing information in a database and automatically categorizing users |
| CN105843381A (zh) * | 2016-03-18 | 2016-08-10 | 北京光年无限科技有限公司 | 用于实现多模态交互的数据处理方法及多模态交互系统 |
| CN106886820A (zh) * | 2017-02-08 | 2017-06-23 | 深圳市科迈爱康科技有限公司 | 智能信息处理方法及系统 |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8811748B2 (en) * | 2011-05-20 | 2014-08-19 | Autodesk, Inc. | Collaborative feature extraction system for three dimensional datasets |
| US9754210B2 (en) * | 2014-04-01 | 2017-09-05 | Microsoft Technology Licensing, Llc | User interests facilitated by a knowledge base |
| US20160306798A1 (en) * | 2015-04-16 | 2016-10-20 | Microsoft Corporation | Context-sensitive content recommendation using enterprise search and public search |
| CN108764480B (zh) * | 2016-08-23 | 2020-07-07 | 上海智臻智能网络科技股份有限公司 | 一种信息处理的系统 |
-
2017
- 2017-02-08 CN CN201710068552.XA patent/CN106886820A/zh active Pending
- 2017-07-25 WO PCT/CN2017/094348 patent/WO2018145412A1/zh active Application Filing
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5974412A (en) * | 1997-09-24 | 1999-10-26 | Sapient Health Network | Intelligent query system for automatically indexing information in a database and automatically categorizing users |
| CN105843381A (zh) * | 2016-03-18 | 2016-08-10 | 北京光年无限科技有限公司 | 用于实现多模态交互的数据处理方法及多模态交互系统 |
| CN106886820A (zh) * | 2017-02-08 | 2017-06-23 | 深圳市科迈爱康科技有限公司 | 智能信息处理方法及系统 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106886820A (zh) | 2017-06-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| WO2020107765A1 (zh) | 语句分析处理方法、装置、设备以及计算机可读存储介质 | |
| WO2020164281A1 (zh) | 基于文字定位识别的表格解析方法、介质及计算机设备 | |
| WO2018090740A1 (zh) | 一种基于混合现实技术实现陪伴的方法及装置 | |
| WO2019050360A1 (en) | ELECTRONIC DEVICE AND METHOD FOR AUTOMATICALLY SEGMENTING TO BE HUMAN IN AN IMAGE | |
| WO2013157848A1 (en) | Method of displaying multimedia exercise content based on exercise amount and multimedia apparatus applying the same | |
| WO2020107761A1 (zh) | 广告文案处理方法、装置、设备及计算机可读存储介质 | |
| WO2016107231A1 (zh) | 一种3d场景中输入手势的系统和方法 | |
| WO2016111573A1 (ko) | 파일 변환 방법 및 장치 | |
| WO2015065006A1 (en) | Multimedia apparatus, online education system, and method for providing education content thereof | |
| WO2018145412A1 (zh) | 智能信息处理方法及系统 | |
| WO2020235804A1 (ko) | 포즈 유사도 판별 모델 생성방법 및 포즈 유사도 판별 모델 생성장치 | |
| WO2020107762A1 (zh) | Ctr预估方法、装置及计算机可读存储介质 | |
| WO2020119069A1 (zh) | 基于自编码神经网络的文本生成方法、装置、终端及介质 | |
| WO2020207038A1 (zh) | 基于人脸识别的人数统计方法、装置、设备及存储介质 | |
| WO2022255529A1 (ko) | 머신 러닝 기반의 립싱크 영상 생성을 위한 학습 방법 및 이를 수행하기 위한 립싱크 영상 생성 장치 | |
| WO2021012508A1 (zh) | Ai影像识别方法、装置、设备及存储介质 | |
| WO2012034469A1 (zh) | 基于手势的人机交互方法及系统、计算机存储介质 | |
| WO2015034131A2 (ko) | 공간 제스처 인식 장치 및 방법 | |
| WO2023074980A1 (ko) | 동작 인식 기반 상호작용 방법 및 기록 매체 | |
| WO2015196878A1 (zh) | 一种电视虚拟触控方法及系统 | |
| WO2021017332A1 (zh) | 语音控制报错方法、电器及计算机可读存储介质 | |
| WO2023224433A1 (en) | Information generation method and device | |
| WO2016036049A1 (ko) | 검색 서비스 제공 장치, 시스템, 방법 및 컴퓨터 프로그램 | |
| WO2022004970A1 (ko) | 신경망 기반의 특징점 학습 장치 및 방법 | |
| WO2019216484A1 (ko) | 전자 장치 및 그 동작방법 |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| 121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 17896055 Country of ref document: EP Kind code of ref document: A1 |
|
| NENP | Non-entry into the national phase |
Ref country code: DE |
|
| 122 | Ep: pct application non-entry in european phase |
Ref document number: 17896055 Country of ref document: EP Kind code of ref document: A1 |