Intelligent Input Method and System Based on Deep Reinforcement Learning

Abstract

The present invention provides an intelligent input method and system based on deep reinforcement learning, which can select a text input habit database that matches the current user individual performing text input, thereby improving the personalization degree and text input experience of different users during text input. In addition, it can also associate and find several appropriate candidate texts from the selected text input habit database according to the overall meaning of the sentence corresponding to the user during the current text input process for the user to select for input, thus effectively improving the speed and accuracy of the user's text input.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent text input, and particularly to an intelligent input method and system based on deep reinforcement learning. Background Art

[0002] Currently, all text input methods have the function of memorizing text input habits, which can record the text input habits of users on corresponding terminal devices, so as to facilitate the quick push of the next possible text input by the user during the next text input process, thereby improving the text input speed and accuracy of the user. However, the text input methods of the prior art can only construct corresponding text input habit databases based on the historical text input data of the same terminal device, and cannot form a suitable text input habit database for different user individuals, which seriously reduces the experience of different users using text input methods for text input. It can be seen that the prior art needs an intelligent text input mode that can determine a matching text input habit database for different users themselves, and can associate and push appropriate texts according to the user's real-time text input situation. Summary of the Invention

[0003] In view of the defects existing in the prior art, the present invention provides an intelligent input method and system based on deep reinforcement learning. By photographing the user who is currently inputting text, the facial image information of the user is obtained, and the facial image information is preprocessed. Then, according to the preprocessed facial image information, the text input habit database corresponding to the user is determined. According to the text input habit database and the overall semantic meaning of the sentence corresponding to the user's current text input, the text appearance probability information of a certain text included in the text input habit database after the user inputs the current text is determined. Then, according to the text appearance probability information, several appropriate candidate texts are screened from the text input habit database, and the several candidate texts are sequentially displayed in the candidate text area of the corresponding text input box. It can be seen that the intelligent input method and system based on deep reinforcement learning can select a matching text input habit database for the user individual who is currently inputting text, which can improve the personalization degree and text input experience of different users. In addition, according to the overall meaning of the sentence corresponding to the user during the current text input process, several appropriate candidate texts can be associated and found from the selected text input habit database for the user to select for input, thereby effectively improving the speed and accuracy of the user's text input.

[0004] The present invention provides an intelligent input method based on deep reinforcement learning, which is characterized by including the following steps:

[0005] Step S1, photograph the user who is currently inputting text to obtain the facial image information of the user, preprocess the facial image information, and then determine the text input habit database corresponding to the user according to the preprocessed facial image information;

[0006] Step S2, determine the text appearance probability information of a certain text included in the text input habit database when the user inputs the current text according to the text input habit database and the overall semantic meaning of the statement corresponding to the user's current text input;

[0007] Step S3, screen a number of appropriate candidate texts from the text input habit database according to the text appearance probability information, and sequentially display the number of candidate texts in the candidate text area of the corresponding text input box;

[0008] Further, in the step S1, photograph the user who is currently inputting text to obtain the facial image information of the user, preprocess the facial image information, and then determine the text input habit database corresponding to the user specifically includes:

[0009] Step S101, perform binocular photographing on the facial area of the user who is currently inputting text to obtain a first perspective image and a second perspective image of the user, and determine the facial image information of the user according to the parallax between the first perspective image and the second perspective image;

[0010] Step S102, perform background noise filtering processing and pixel smoothing processing on the facial image information to obtain the preprocessed facial image information;

[0011] Step S103, determine the matching degree evaluation value between the preprocessed facial image information and a number of text input habit databases according to the following formula (1):

[0012]

[0013] In the above formula (1), Y t represents the matching degree evaluation value between the preprocessed facial image information and the t-th text input habit database, and A ab represents the chromaticity value corresponding to the pixel point at the a-th row and b-th column of the facial image in the preprocessed facial image information, It represents the chromaticity value corresponding to the pixel at the a-th row and b-th column of the face image in the standard face image information corresponding to the t-th text input habit database. h represents the total number of pixel rows in the face image in the preprocessed face image information or the face image in the standard face image information, and l represents the total number of pixel columns in the face image in the preprocessed face image information or the face image in the standard face image information;

[0014] Step S104: From the several matching degree evaluation values obtained in the above step S103, determine the matching degree evaluation value with the maximum value, and use the text input habit database corresponding to the matching degree evaluation value with the maximum value as the text input habit database corresponding to the user;

[0015] Further, in the above step S2, according to the text input habit database and the overall semantic meaning of the statement corresponding to the user's current text input, determining the text occurrence probability information of a certain text included in the text input habit database after the user inputs the current text specifically includes:

[0016] Step S201: According to the text input habit database and the following formula (2), determine the text occurrence probability value of a certain text included in the text input habit database after the user inputs the current text:

[0017]

[0018] In the above formula (2), P r represents the text occurrence probability value of the r-th text included in the text input habit database after the user inputs the i-th text, represents the number of times the i-th text in the j-th statement input by the user is adjacent to the r-th text included in the text input habit database, g i represents the number of times the i-th text appears in the user's historical input process, m represents the total number of statements in the user's historical input, and S represents the total number of texts included in the text input habit database;

[0019] Step S202: According to the text occurrence probability value and the following formula (3), determine the text occurrence probability value of a certain text included in the text input habit database after the user inputs the current text under the limitation of the overall semantic meaning of the statement:

[0020]

[0021] In the above formula (3), T rIt represents the probability value of the occurrence of the r-th character included in the character input habit database after the user inputs the i-th character under the overall semantic limitation of the statement. P1 represents the probability value of the occurrence of the 1st character included in the character input habit database after the user inputs the i-th character; P r It represents the probability value of the occurrence of the r-th character included in the character input habit database after the user inputs the i-th character. P r+1 It represents the probability value of the occurrence of the (r + 1)-th character included in the character input habit database after the user inputs the i-th character. It represents the number of times the i-th character in the j-th statement of the user's historical input is adjacent to the r-th character included in the character input habit database. It represents the number of times the i-th character in the 1st statement of the user's historical input is adjacent to the r-th character included in the character input habit database. m represents the total number of statements in the user's historical input, n represents the total number of characters that the user has completed inputting during the process of editing the current statement, and S represents the total number of characters included in the character input habit database;

[0022] Furthermore, in the step S3, according to the character occurrence probability information, screening a number of appropriate candidate characters from the character input habit database and sequentially displaying the number of candidate characters on the corresponding input interface specifically includes:

[0023] Step S301, according to the probability value T of the occurrence of the r-th character included in the character input habit database after the user inputs the i-th character determined above under the overall semantic limitation of the statement r , screening a number of corresponding candidate characters from the character input habit database;

[0024] Step S302, in the order from large to small corresponding to the character occurrence probability value T of the number of candidate characters r , sequentially displaying the number of candidate characters on the corresponding input interface.

[0025] The present invention also provides an intelligent input system based on deep reinforcement learning, which is characterized in that it includes a face image shooting and preprocessing module, a character input habit database determination module, a character occurrence probability information determination module, and a candidate character determination module; wherein,

[0026] The face image shooting and preprocessing module is used to shoot the user who is currently inputting characters, thereby obtaining the face image information of the user and preprocessing the face image information;

[0027] The text input habit database determination module is used to determine the text input habit database corresponding to the user according to the preprocessed facial image information;

[0028] The text appearance probability information determination module is used to determine the text appearance probability information of a certain text included in the text input habit database after the user inputs the current text according to the text input habit database and the overall semantics of the sentence corresponding to the user's current text input;

[0029] The candidate text determination module is used to screen a number of appropriate candidate texts from the text input habit database according to the text appearance probability information, and sequentially display the number of candidate texts in the candidate text area of the corresponding text input box;

[0030] Furthermore, the facial image shooting and preprocessing module shoots the user who is currently inputting text to obtain the facial image information of the user, and the specific preprocessing of the facial image information includes:

[0031] Perform binocular shooting on the facial area of the user who is currently inputting text, so as to obtain a first perspective image and a second perspective image of the user, and determine the facial image information of the user according to the parallax between the first perspective image and the second perspective image;

[0032] Then perform background noise filtering processing and pixel smoothing processing on the facial image information to obtain the preprocessed facial image information;

[0033] And,

[0034] The specific process by which the text input habit database determination module determines the text input habit database corresponding to the user according to the preprocessed facial image information includes:

[0035] According to the following formula (1), determine the matching degree evaluation value between the preprocessed facial image information and a number of text input habit databases:

[0036]

[0037] In the above formula (1), Y t represents the matching degree evaluation value between the preprocessed facial image information and the t-th text input habit database, and A ab represents the chromaticity value corresponding to the pixel point in the a-th row and b-th column of the facial image in the preprocessed facial image information, Denote the chroma value corresponding to the pixel at the \(a\)-th row and \(b\)-th column of the face image in the standard face image information corresponding to the \(t\)-th text input habit database. \(h\) represents the total number of pixel rows in the face image in the preprocessed face image information or the face image in the standard face image information, and \(l\) represents the total number of pixel columns in the face image in the preprocessed face image information or the face image in the standard face image information;

[0038] Then, from the obtained several matching degree evaluation values, determine the matching degree evaluation value with the maximum value, and use the text input habit database corresponding to the matching degree evaluation value with the maximum value as the text input habit database corresponding to the user;

[0039] Further, the text occurrence probability information determination module determines the text occurrence probability information that a certain text included in the text input habit database appears after the user inputs the current text according to the text input habit database and the overall semantics of the statement corresponding to the user's current text input, specifically including:

[0040] According to the text input habit database and the following formula (2), determine the text occurrence probability value that a certain text included in the text input habit database appears after the user inputs the current text:

[0041]

[0042] In the above formula (2), \(P\) r represents the text occurrence probability value that the \(r\)-th text included in the text input habit database appears after the user inputs the \(i\)-th text. represents the number of times the \(i\)-th text in the \(j\)-th statement input by the user is adjacent to the \(r\)-th text included in the text input habit database. \(g\) i represents the number of times the \(i\)-th text appears in the user's historical input process. \(m\) represents the total number of statements input by the user historically, and \(S\) represents the total number of texts included in the text input habit database;

[0043] Then, according to the text occurrence probability value and the following formula (3), determine the text occurrence probability value that a certain text included in the text input habit database appears after the user inputs the current text under the limitation of the overall semantics of the statement:

[0044]

[0045] In the above formula (3), \(T\) r represents the text occurrence probability value that the \(r\)-th text included in the text input habit database appears after the user inputs the \(i\)-th text under the limitation of the overall semantics of the statement. \(P\) rIndicates the probability value of the occurrence of the r-th character included in the character input habit database after the user inputs the i-th character, P r+1 Indicates the probability value of the occurrence of the (r + 1)-th character included in the character input habit database after the user inputs the i-th character Indicates the number of times the i-th character in the j-th sentence of the user's historical input is adjacent to the r-th character included in the character input habit database Indicates the number of times the i-th character in the 1st sentence of the user's historical input is adjacent to the r-th character included in the character input habit database. m represents the total number of sentences in the user's historical input, n represents the total number of characters that the user has completed inputting during the process of editing the current sentence, and S represents the total number of characters included in the character input habit database;

[0046] Further, the candidate character determination module screens a number of appropriate candidate characters from the character input habit database according to the character occurrence probability information, and sequentially displays the number of candidate characters on the corresponding input interface, specifically including:

[0047] According to the probability value T of the occurrence of the r-th character included in the character input habit database after the user inputs the i-th character determined as above under the overall semantic limitation of the sentence r , screen a number of corresponding candidate characters from the character input habit database;

[0048] Then, in the order from large to small corresponding to the probability value T of the occurrence of the number of candidate characters r Sequentially display the number of candidate characters on the corresponding input interface.

[0049] Compared with the prior art, the intelligent input method and system based on deep reinforcement learning photograph the user of the current input text to obtain the facial image information of the user, preprocess the facial image information, and then determine the text input habit database corresponding to the user according to the preprocessed facial image information. According to the text input habit database and the overall semantic meaning of the sentence corresponding to the user's current text input, determine the text appearance probability information of a certain text included in the text input habit database after the user inputs the current text. Then, according to the text appearance probability information, screen a number of appropriate candidate texts from the text input habit database and display the candidate texts in the candidate text area of the corresponding text input box in sequence. It can be seen that the intelligent input method and system based on deep reinforcement learning can select a text input habit database that matches the individual user who is currently inputting text, which can improve the personalization degree and text input experience of different users when inputting text. In addition, it can also associate and find a number of appropriate candidate texts from the selected text input habit database according to the overall meaning of the sentence corresponding to the user during the current text input process for the user to select for input, thereby effectively improving the speed and accuracy of the user's text input.

[0050] Other features and advantages of the present invention will be described in the following specification, and, in part, will be obvious from the specification, or will be understood by implementing the present invention. The objectives and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written specification, claims, and drawings.

[0051] The technical solutions of the present invention will be further described in detail below with reference to the drawings and embodiments. Brief Description of the Drawings

[0052] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0053] Figure 1 It is a schematic flowchart of the intelligent input method based on deep reinforcement learning provided by the present invention.

[0054] Figure 2 It is a schematic structural diagram of the intelligent input system based on deep reinforcement learning provided by the present invention. Detailed Embodiments

[0055] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0056] Refer to Figure 1 , which is a schematic flowchart of an intelligent input method based on deep reinforcement learning provided by an embodiment of the present invention. The intelligent input method based on deep reinforcement learning includes the following steps:

[0057] Step S1, photograph the user who is currently inputting text to obtain the facial image information of the user, preprocess the facial image information, and then determine the text input habit database corresponding to the user according to the preprocessed facial image information;

[0058] Step S2, determine the text appearance probability information of a certain text included in the text input habit database when the text appears after the user inputs the current text according to the text input habit database and the overall semantics of the sentence corresponding to the user's current text input;

[0059] Step S3, screen a number of appropriate candidate texts from the text input habit database according to the text appearance probability information, and sequentially display the number of candidate texts in the candidate text area of the corresponding text input box.

[0060] The beneficial effects of the above technical solutions are as follows: The intelligent input method based on deep reinforcement learning can select a text input habit database that matches the user individual who is currently inputting text, which can improve the personalization degree and text input experience of different users when inputting text. In addition, it can also associate and find a number of appropriate candidate texts from the selected text input habit database according to the overall meaning of the sentence corresponding to the user during the current text input for the user to select for input, thereby effectively improving the speed and accuracy of the user's text input.

[0061] Preferably, in step S1, photographing the user who is currently inputting text to obtain the facial image information of the user, preprocessing the facial image information, and then determining the text input habit database corresponding to the user specifically includes:

[0062] Step S101, binocularly photograph the facial area of the user who is currently inputting text to obtain a first perspective image and a second perspective image of the user, and determine the facial image information of the user according to the parallax between the first perspective image and the second perspective image;

[0063] Step S102: Perform background noise filtering and pixel smoothing on the facial image information to obtain the preprocessed facial image information.

[0064] Step S103: According to the following formula (1), determine the matching degree evaluation value between the preprocessed facial image information and several text input habit databases:

[0065]

[0066] In the above formula (1), Y t represents the matching degree evaluation value between the preprocessed facial image information and the t-th text input habit database, and A ab represents the chromaticity value corresponding to the pixel point at the a-th row and b-th column of the facial image in the preprocessed facial image information. represents the chromaticity value corresponding to the pixel point at the a-th row and b-th column of the facial image in the standard facial image information corresponding to the t-th text input habit database. h represents the total number of pixel rows in the facial image in the preprocessed facial image information or the facial image in the standard facial image information, and l represents the total number of pixel columns in the facial image in the preprocessed facial image information or the facial image in the standard facial image information.

[0067] Step S104: From the several matching degree evaluation values obtained in the above step S103, determine the matching degree evaluation value with the maximum value, and use the text input habit database corresponding to the matching degree evaluation value with the maximum value as the text input habit database corresponding to the user.

[0068] The beneficial effects of the above technical solution are as follows: By photographing the facial image of the user currently performing text input, and performing noise reduction preprocessing and image recognition processing on the photographed facial image, the text input habit database that matches the user can be found in several different text input habit databases, which can facilitate the construction of multiple different text input habit databases for different users on the same terminal device, and accurately and quickly call the matching text input habit database for subsequent text input association and push when needed by the user, thereby improving the recording effectiveness and compatibility of the terminal device with the text input habits of different users.

[0069] Preferably, in this step S2, according to the text input habit database and the overall semantic meaning of the statement corresponding to the user's current text input, determining the text appearance probability information of a certain text included in the text input habit database after the user inputs the current text specifically includes:

[0070] Step S201: According to the text input habit database and the following formula (2), determine the text occurrence probability value of a certain text included in the text input habit database after the user inputs the current text:

[0071]

[0072] In the above formula (2), P r represents the text occurrence probability value of the r-th text included in the text input habit database after the user inputs the i-th text, represents the number of times the i-th text in the j-th sentence of the user's historical input is adjacent to the r-th text included in the text input habit database, g i represents the number of times the i-th text appears in the user's historical input process, m represents the total number of sentences in the user's historical input, and S represents the total number of texts included in the text input habit database;

[0073] Step S202: According to the text occurrence probability value and the following formula (3), determine the text occurrence probability value of a certain text included in the text input habit database after the user inputs the current text under the overall semantic limitation of the sentence:

[0074]

[0075] In the above formula (3), T r represents the text occurrence probability value of the r-th text included in the text input habit database after the user inputs the i-th text under the overall semantic limitation of the sentence, P r represents the text occurrence probability value of the r-th text included in the text input habit database after the user inputs the i-th text, P r+1 represents the text occurrence probability value of the (r + 1)-th text included in the text input habit database after the user inputs the i-th text, represents the number of times the i-th text in the j-th sentence of the user's historical input is adjacent to the r-th text included in the text input habit database, represents the number of times the i-th text in the 1st sentence of the user's historical input is adjacent to the r-th text included in the text input habit database, m represents the total number of sentences in the user's historical input, n represents the total number of texts that the user has completed inputting during the editing of the current sentence, and S represents the total number of texts included in the text input habit database.

[0076] The beneficial effects of the above technical solution are as follows: By successively determining the probability value of the appearance of a certain character included in the character input habit database after the user inputs the current character during the current character input process, and under the overall semantic limitation of the sentence that the user needs to input, the probability value of the appearance of a certain character included in the character input habit database after the user inputs the current character, it is possible to gradually determine the character that the user may need to input next from two different levels: the single-character / word input level and the entire sentence input level, thereby further narrowing the scope of the text association and push that the user may need to input next, and thus maximizing the determination accuracy and reliability of the character that the user may need to input next.

[0077] Preferably, in step S3, according to the character appearance probability information, screening a number of appropriate candidate characters from the character input habit database and successively displaying the number of candidate characters on the corresponding input interface specifically includes:

[0078] Step S301, according to the determined probability value T of the appearance of the r-th character included in the character input habit database after the user inputs the i-th character under the overall semantic limitation of the sentence r , screening a corresponding number of candidate characters from the character input habit database;

[0079] Step S302, arranging the number of candidate characters in descending order according to the probability value T of the appearance of the number of candidate characters r corresponding thereto, and successively displaying the number of candidate characters in the candidate character area of the corresponding text input box.

[0080] The beneficial effects of the above technical solution are as follows: By arranging the number of candidate characters in descending order according to the probability value T of the appearance of the number of candidate characters r corresponding thereto, and successively displaying the number of candidate characters on the corresponding input interface, it is convenient for the user to quickly and accurately select the next character that the user actually needs to input during the character input process, thereby effectively improving the convenience and usability of the character input.

[0081] Refer to Figure 2 , which is a schematic structural diagram of the intelligent input system based on deep reinforcement learning provided by the embodiment of the present invention. The intelligent input system based on deep reinforcement learning includes a face image shooting and preprocessing module, a character input habit database determination module, a character appearance probability information determination module, and a candidate character determination module; wherein,

[0082] The face image shooting and preprocessing module is used to shoot the user who is currently inputting characters, thereby obtaining the face image information of the user, and preprocessing the face image information;

[0083] The text input habit database determination module is used to determine the text input habit database corresponding to the user according to the preprocessed facial image information;

[0084] The text appearance probability information determination module is used to determine the text appearance probability information of a certain text included in the text input habit database after the user inputs the current text according to the text input habit database and the overall semantics of the statement corresponding to the user's current text input;

[0085] The candidate text determination module is used to screen a number of appropriate candidate texts from the text input habit database according to the text appearance probability information, and display the number of candidate texts on the corresponding input interface in sequence.

[0086] The beneficial effects of the above technical solution are as follows: The intelligent input system based on deep reinforcement learning can select a text input habit database that matches the current user who is inputting text, which can improve the personalization degree and text input experience of different users when inputting text. In addition, it can also associate and find a number of appropriate candidate texts from the selected text input habit database according to the overall meaning of the statement corresponding to the user during the current text input process for the user to select for input, thereby effectively improving the speed and accuracy of the user's text input.

[0087] Preferably, the facial image shooting and preprocessing module shoots the user who is currently inputting text to obtain the facial image information of the user, and the preprocessing of the facial image information specifically includes:

[0088] Perform binocular shooting on the facial area of the user who is currently inputting text to obtain a first perspective image and a second perspective image of the user, and determine the facial image information of the user according to the parallax between the first perspective image and the second perspective image;

[0089] Then perform background noise filtering processing and pixel smoothing processing on the facial image information to obtain the preprocessed facial image information;

[0090] And,

[0091] The text input habit database determination module determines the text input habit database corresponding to the user according to the preprocessed facial image information, specifically including:

[0092] According to the following formula (1), determine the matching degree evaluation value between the preprocessed facial image information and a number of text input habit databases:

[0093]

[0094] In the above formula (1), Yt Denotes the matching degree evaluation value between the pre - processed facial image information and the t - th text input habit database, A ab Denotes the chromaticity value corresponding to the pixel point at the a - th row and b - th column of the facial image in the pre - processed facial image information Denotes the chromaticity value corresponding to the pixel point at the a - th row and b - th column of the facial image in the standard facial image information corresponding to the t - th text input habit database. h represents the total number of pixel rows in the facial image of the pre - processed facial image information or the facial image in the standard facial image information, and l represents the total number of pixel columns in the facial image of the pre - processed facial image information or the facial image in the standard facial image information

[0095] Then, from the several obtained matching degree evaluation values above, determine the matching degree evaluation value with the maximum value, and use the text input habit database corresponding to the matching degree evaluation value with the maximum value as the text input habit database corresponding to this user

[0096] The beneficial effects of the above - mentioned technical solution are as follows: By taking a facial image of the user currently performing text input, and performing noise reduction pre - processing and image recognition processing on the captured facial image, so as to find the text input habit database that matches this user among several different text input habit databases. This can facilitate the construction of multiple different text input habit databases for different users on the same terminal device, and accurately and quickly call the matching text input habit database for subsequent text input association and push when the user needs it, thereby improving the recording effectiveness and compatibility of the terminal device with different users' text input habits

[0097] Preferably, the text appearance probability information determination module determines the text appearance probability information that a certain text included in the text input habit database appears after the user inputs the current text according to the text input habit database and the overall semantics of the statement corresponding to the user's current text input, specifically including:

[0098] According to the text input habit database and the following formula (2), determine the text appearance probability value that a certain text included in the text input habit database appears after the user inputs the current text:

[0099]

[0100] In the above formula (2), P r Denotes the text appearance probability value that the r - th text included in the text input habit database appears after the user inputs the i - th text Denotes the number of times the i - th text in the j - th statement of the user's historical input is adjacent to the r - th text included in the text input habit database, gi represents the number of occurrences of the i-th character in the user's historical input process, m represents the total number of statements in the user's historical input, and S represents the total number of characters included in the character input habit database;

[0101] Then, according to the character occurrence probability value and the following formula (3), determine the character occurrence probability value of a certain character included in the character input habit database when the user inputs the current character under the overall semantic limitation of the statement:

[0102]

[0103] In the above formula (3), T r represents the character occurrence probability value of the r-th character included in the character input habit database when the user inputs the i-th character under the overall semantic limitation of the statement, and P r represents the character occurrence probability value of the r-th character included in the character input habit database after the user inputs the i-th character, and P r+1 represents the character occurrence probability value of the (r + 1)-th character included in the character input habit database after the user inputs the i-th character, represents the number of times the i-th character in the j-th statement of the user's historical input is adjacent to the r-th character included in the character input habit database, represents the number of times the i-th character in the 1st statement of the user's historical input is adjacent to the r-th character included in the character input habit database, m represents the total number of statements in the user's historical input, n represents the total number of characters that have been input by the user during the process of editing the current statement, and S represents the total number of characters included in the character input habit database.

[0104] The beneficial effects of the above technical solution are as follows: By successively determining the character occurrence probability value of a certain character included in the character input habit database when the user inputs the current character during the current character input process and the character occurrence probability value of a certain character included in the character input habit database when the user inputs the current character under the overall semantic limitation of the statement that the user needs to input, it is possible to gradually determine the characters that the user may need to input next from two different levels: the single-character / word input level and the entire statement input level, thereby further narrowing the range of the characters that may be associated with and pushed for the next input, and thus maximizing the determination accuracy and reliability of the characters that the user may need to input next.

[0105] Preferably, the candidate character determination module screens several appropriate candidate characters from the character input habit database according to the character occurrence probability information, and sequentially displays the several candidate characters on the corresponding input interface, specifically including:

[0106] Under the overall semantic limitation of the statement determined above, the probability value T of the r-th character included in the character input habit database appearing after the user inputs the i-th character r , filter a number of candidate characters corresponding to the character input habit database;

[0107] Then, according to the probability value T of the character appearance of a number of the candidate characters r in the order from large to small, display the number of the candidate characters on the corresponding input interface in sequence.

[0108] The beneficial effect of the above technical solution is that by displaying the number of the candidate characters on the corresponding input interface in the order from large to small according to the probability value T of the character appearance of a number of the candidate characters, it is convenient for the user to quickly and accurately select the next character that the user actually needs to input during the character input process, thereby effectively improving the convenience and usability of the character input. r From the content of the above embodiments, it can be seen that the intelligent input method and system based on deep reinforcement learning obtain the facial image information of the user by photographing the user who is currently inputting characters, preprocess the facial image information, and then determine the character input habit database corresponding to the user according to the preprocessed facial image information. According to the character input habit database and the overall semantics of the statement corresponding to the user's current character input, determine the character appearance probability information of a certain character included in the character input habit database after the user inputs the current character. Then, according to the character appearance probability information, filter a number of appropriate candidate characters from the character input habit database and display the number of the candidate characters in sequence in the candidate character area of the corresponding character input box; it can be seen that the intelligent input method and system based on deep reinforcement learning can select the character input habit database that matches the individual user who is currently inputting characters, which can improve the personalization degree and character input experience of different users. In addition, it can also associate and find a number of appropriate candidate characters from the selected character input habit database according to the overall meaning of the statement corresponding to the user during the current character input process for the user to select for input, thereby effectively improving the speed and accuracy of the user's character input.

[0109] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention also intends to include these changes and modifications.

[0110] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention also intends to include these changes and modifications.

Claims

1. An intelligent input method based on deep reinforcement learning, characterized in that, It includes the following steps: Step S1: Take a photo of the user who is currently inputting text to obtain the facial image information of the user, preprocess the facial image information, and then determine the text input habit database corresponding to the user according to the preprocessed facial image information; Step S2: Determine the text appearance probability information of a certain text included in the text input habit database when the user appears after inputting the current text according to the text input habit database and the overall semantic meaning of the sentence corresponding to the user's current text input; Step S3: Screen a number of appropriate candidate texts from the text input habit database according to the text appearance probability information, and sequentially display the number of candidate texts in the candidate text area of the corresponding text input box; Among them, in the step S1, taking a photo of the user who is currently inputting text to obtain the facial image information of the user, preprocessing the facial image information, and then determining the text input habit database corresponding to the user specifically includes: Step S101: Take binocular photos of the facial area of the user who is currently inputting text to obtain a first perspective image and a second perspective image of the user, and determine the facial image information of the user according to the parallax between the first perspective image and the second perspective image; Step S102: Perform background noise filtering processing and pixel smoothing processing on the facial image information to obtain the preprocessed facial image information; Step S103: Determine the matching degree evaluation value between the preprocessed facial image information and a number of text input habit databases according to the following formula (1): In the above formula (1), Y t represents the matching degree evaluation value between the pre - processed facial image information and the t - th text input habit database, A ab represents the chromaticity value corresponding to the pixel at the a - th row and b - th column of the facial image in the pre - processed facial image information, represents the chromaticity value corresponding to the pixel at the a - th row and b - th column of the facial image in the standard facial image information corresponding to the t - th text input habit database, h represents the total number of pixel rows in the facial image in the pre - processed facial image information or the facial image in the standard facial image information, and l represents the total number of pixel columns in the facial image in the pre - processed facial image information or the facial image in the standard facial image information; Step S104: Determine the matching degree evaluation value with the maximum value from the number of matching degree evaluation values obtained in the above step S103, and use the text input habit database corresponding to the matching degree evaluation value with the maximum value as the text input habit database corresponding to the user; Among them, in the step S2, determining the text appearance probability information of a certain text included in the text input habit database when the user appears after inputting the current text according to the text input habit database and the overall semantic meaning of the sentence corresponding to the user's current text input specifically includes: Step S201: Determine the text appearance probability value of a certain text included in the text input habit database when the user appears after inputting the current text according to the text input habit database and the following formula (2); In the above formula (2), P r represents the probability value of the occurrence of the r-th word included in the word input habit database after the user inputs the i-th word, represents the number of times the i-th word in the j-th sentence of the user's historical input is adjacent to the r-th word included in the word input habit database, g i represents the number of occurrences of the i-th word in the user's historical input process, m represents the total number of sentences in the user's historical input, and S represents the total number of words included in the word input habit database; Step S202: Determine the text appearance probability value of a certain text included in the text input habit database when the user appears after inputting the current text under the limitation of the overall semantic meaning of the sentence according to the text appearance probability value and the following formula (3); In the above formula (3), T r represents the probability value of the r-th word included in the word input habit database appearing after the user inputs the i-th word under the overall semantic limitation of the statement. P1 represents the probability value of the 1st word included in the word input habit database appearing after the user inputs the i-th word; P r represents the probability value of the r-th word included in the word input habit database appearing after the user inputs the i-th word, P r+1 represents the probability value of the (r + 1)-th word included in the word input habit database appearing after the user inputs the i-th word, represents the number of times the i-th word in the j-th statement input by the user is adjacent to the r-th word included in the word input habit database, represents the number of times the i-th word in the 1st statement input by the user is adjacent to the r-th word included in the word input habit database. m represents the total number of statements input by the user, n represents the total number of words that have been input by the user during the process of editing the current statement, and S represents the total number of words included in the word input habit database.

2. The intelligent input method based on deep reinforcement learning according to claim 1, characterized in that: In the step S3, screening a number of appropriate candidate texts from the text input habit database according to the text appearance probability information and sequentially displaying the number of candidate texts in the corresponding input interface specifically includes: Step S301, according to the determined word occurrence probability value T that after the user inputs the i-th word, the r-th word included in the word input habit database appears under the overall semantic limitation of the statement r , screen a number of candidate words corresponding thereto from the word input habit database; Step S302, display a number of the candidate words in the corresponding input interface in sequence in the order from large to small according to the word appearance probability value T of the number of the candidate words r corresponding thereto 3. An intelligent input system based on deep reinforcement learning, characterized in that, It includes a facial image capture and preprocessing module, a text input habit database determination module, a text appearance probability information determination module, and a candidate text determination module; among them, the facial image capture and preprocessing module is used to capture the user who is currently inputting text, so as to obtain the facial image information of the user, and preprocess the facial image information; the text input habit database determination module is used to determine the text input habit database corresponding to the user according to the preprocessed facial image information; the text appearance probability information determination module is used to determine the text appearance probability information of a certain text included in the text input habit database when the user inputs the current text according to the text input habit database and the overall semantics of the sentence corresponding to the user's current text input; the candidate text determination module is used to screen a number of appropriate candidate texts from the text input habit database according to the text appearance probability information, and display the number of candidate texts in the candidate text area of the corresponding text input box in sequence; among them, the facial image capture and preprocessing module captures the user who is currently inputting text, so as to obtain the facial image information of the user, and the specific preprocessing of the facial image information includes: performing binocular capture on the facial area of the user who is currently inputting text, so as to obtain a first perspective image and a second perspective image of the user, and determining the facial image information of the user according to the parallax between the first perspective image and the second perspective image; then performing background noise filtering processing and pixel smoothing processing on the facial image information, so as to obtain the preprocessed facial image information; and, the specific process of the text input habit database determination module determining the text input habit database corresponding to the user according to the preprocessed facial image information includes: determining the matching degree evaluation value between the preprocessed facial image information and a number of text input habit databases according to the following formula (1): In the above formula (1), Y t represents the matching degree evaluation value between the preprocessed facial image information and the t-th text input habit database, and A ab represents the chromaticity value corresponding to the pixel at the a-th row and b-th column of the facial image in the preprocessed facial image information. represents the chromaticity value corresponding to the pixel at the a-th row and b-th column of the facial image in the standard facial image information corresponding to the t-th text input habit database. h represents the total number of pixel rows in the facial image in the preprocessed facial image information or the facial image in the standard facial image information, and l represents the total number of pixel columns in the facial image in the preprocessed facial image information or the facial image in the standard facial image information; then determining the matching degree evaluation value with the maximum value from the above-obtained number of matching degree evaluation values, and using the text input habit database corresponding to the matching degree evaluation value with the maximum value as the text input habit database corresponding to the user; among them, the specific process of the text appearance probability information determination module determining the text appearance probability information of a certain text included in the text input habit database when the user inputs the current text according to the text input habit database and the overall semantics of the sentence corresponding to the user's current text input includes: determining the text appearance probability value of a certain text included in the text input habit database when the user inputs the current text according to the text input habit database and the following formula (2): In the above formula (2), P r represents the probability value of the occurrence of the r-th word included in the word input habit database after the user inputs the i-th word, represents the number of times the i-th word in the j-th sentence of the user's historical input is adjacent to the r-th word included in the word input habit database, g i represents the number of times the i-th word appears during the user's historical input process, m represents the total number of sentences in the user's historical input, and S represents the total number of words included in the word input habit database; then determining the text appearance probability value of a certain text included in the text input habit database when the user inputs the current text under the limitation of the overall semantics of the sentence according to the text appearance probability value and the following formula (3): In the above formula (3), T r represents the probability value of the occurrence of the r-th character included in the character input habit database after the user inputs the i-th character under the overall semantic limitation of the statement, P r represents the probability value of the occurrence of the r-th character included in the character input habit database after the user inputs the i-th character, P r+1 represents the probability value of the occurrence of the (r + 1)-th character included in the character input habit database after the user inputs the i-th character, represents the number of times the i-th character in the j-th statement of the user's historical input is adjacent to the r-th character included in the character input habit database, represents the number of times the i-th character in the 1st statement of the user's historical input is adjacent to the r-th character included in the character input habit database, m represents the total number of statements of the user's historical input, n represents the total number of characters that the user has completed inputting during the process of editing the current statement, and S represents the total number of characters included in the character input habit database.

4. The intelligent input system based on deep reinforcement learning according to claim 3, wherein: The to-be-selected text determination module screens a number of appropriate to-be-selected texts from the text input habit database according to the text appearance probability information, and sequentially displays the number of to-be-selected texts on the corresponding input interface, which specifically includes: Under the overall semantic limitation of the statement determined above, the probability value T of the r-th character in the character input habit database appearing after the user inputs the i-th character r , and screen a number of candidate characters corresponding thereto from the character input habit database; Then, according to the text occurrence probability values T of several said candidate texts r in the corresponding order from large to small, display several said candidate texts on the corresponding input interface in sequence.

Images