System capable of inputting characters through voice

Abstract

The invention discloses a system capable of inputting characters through voice, and relates to the field of voice input systems.The system comprises a bracelet, a voice processing module, a feature input module, a correction module and a deep learning module, the feature input module, the correction module and the deep learning module are all installed in the bracelet, and a sound receiver and a loudspeaker box are fixedly connected to the interior of the bracelet through bolts; and a touch screen is embedded in the surface of the bracelet. The voice processing module is used for performing voice interaction and text processing and forming a voice input record; the feature input module is used for inputting voiceprint features, mouth shape features, facial features and gesture instructions and forming a private account according to the input voiceprint features, mouth shape features, facial features and gesture instructions; the feature input module comprises a voiceprint feature input unit, a mouth shape feature input unit, a gesture input unit and a private account input unit. The method is complete in logic, the pertinence of the system can be effectively improved, and meanwhile protection on user privacy is improved.

Description

Technical Field

[0001] The invention relates to the field of voice input systems, in particular to a voice input text system. Background Art

[0002] Speech recognition is a technology that converts human speech into text. It uses multidisciplinary knowledge such as computer science, linguistics, and acoustics to analyze and process input speech signals through algorithms and models, and ultimately outputs the corresponding text content.

[0003] The existing speech recognition system mainly includes speech extraction, acoustic model, language model and text output module, which can effectively collect and process speech and output it in the form of text. However, it lacks specificity for different users and thus lacks privacy protection for users.

[0004] Therefore, the present invention provides a voice input text system, which can associate a preset fixed text according to several short words and sentences input by the user's voice, thereby improving the voice input speed; it can effectively improve the pertinence of the system and improve the protection of user privacy. Summary of the invention

[0005] In order to solve the above problems, the present invention provides a text input system that can be used for voice input. Through the design of a feature entry module and a deep learning module, the user's voiceprint features are fully understood, and a private account is formed based on the voiceprint features, thereby improving the system's pertinence and privacy.

[0006] In order to achieve the above-mentioned object, the technical solution of the present invention is as follows: a text input system capable of voice input includes a bracelet, a voice processing module, a feature entry module, a correction module and a deep learning module, wherein the feature entry module, the correction module and the deep learning module are all installed in the bracelet;

[0007] A speech processing module is used for speech interaction and text processing, and forms a speech input record; the speech processing module is input with a language database;

[0008] A feature input module, used to input voiceprint features, lip shape features, facial features and gesture commands, and to form a private account based on the input voiceprint features, lip shape features, facial features and gesture commands;

[0009] A correction module is used to correct the text and form a correction record;

[0010] The deep learning module is used to optimize the system according to the user's voice input habits, interaction context and speaking style, and analyze the voice input records and correction records of private accounts to form an input habit library and associated vocabulary library of private accounts; and transmit the relevant input habits and associated vocabulary to the voice processing module during the text processing process.

[0011] The following principles and beneficial effects are achieved by adopting the above scheme:

[0012] The principle of the basic solution is: when the user conducts voice interaction for the first time, the voice processing module receives the language spoken by the user, identifies the language and dialect of the language spoken by the user, compares the language with the language database, retrieves the corresponding text, displays the text, and outputs the output text in voice using the corresponding language and dialect.

[0013] During the initial interaction, the voice processing module will guide the user to create the account name the user wants and set the user's private account password. After the setting is completed, the voice processing module will guide the user to enter the specified verification code through voice. During this process, the feature entry module records the user's voiceprint features, lip shape features and facial features. After the feature entry is completed, the user's private account is formed in the system. The user can enter the shortcut gesture instructions he needs through the gesture entry unit. For example, gesture 1 represents the user's name. When the user makes gesture 1 with his hands, the voice processing module outputs the user's name.

[0014] During daily use, the voice processing module estimates the user's voice input each time to form a voice input record. At the same time, if the user's voice input is wrong when the text is output, the user can adjust the output text through the correction module, replace the wrong text, sentence or punctuation, or adjust the text format; the correction module will form a correction record based on the user's adjustment;

[0015] In daily interactions, the deep learning module will receive the correction records transmitted by the correction module and the voice input records of the voice processing module, analyze the user's commonly used keywords and sentences, and analyze the user's pronunciation in combination with the user's mouth changes, and compare the user's pronunciation with the output text, thereby assisting the voice processing module to output text more accurately; at the same time, the deep learning module will judge the user's emotions and context of speech based on changes in the user's facial expressions, further improving the accuracy of text and punctuation during text output.

[0016] The beneficial effects of the basic scheme are: 1. By integrating the speech processing module, the feature entry module, the correction module and the deep learning module, the present invention can continuously optimize the user's speech input experience. The speech processing module uses the language database to perform preliminary text conversion, while the deep learning module continuously adjusts and optimizes the recognition algorithm according to the user's input habits and correction records, thereby improving the accuracy of speech-to-text conversion. At the same time, the deep learning module can judge the user's speaking emotions and context by analyzing the user's lip shape changes and facial expressions, so as to more accurately add punctuation or adjust the tone when outputting text, so that the output text is more in line with the user's true intentions and emotional expressions. The present invention has the ability of self-learning and optimization. The deep learning module can continuously update and improve the user's input habit library and associated vocabulary library by continuously analyzing the user's input habits and correction records, thereby providing more accurate and personalized services in future interactions.

[0017] 2. The present invention records the user's voiceprint features, lip shape features, facial features and gesture instructions through the feature recording module to form a private account. This not only enhances the security of the account, but also enables the system to better understand the user's personalized needs.

[0018] 3. Users can quickly enter commonly used information, such as name, address, and specific text such as shortcut words, by setting shortcut gesture commands, which greatly improves the convenience of voice interaction.

[0019] Furthermore, the language database includes dialects and scripts of English, Chinese, French, Russian, Spanish and Arabic.

[0020] The beneficial effect of the basic solution is that by including multiple major international languages ​​and their dialects, the present invention has the ability to support globalization. This allows users to achieve efficient voice input and interaction through the present invention no matter where they are and what language they use, greatly broadening the scope of application and user groups of the system. The system can intelligently identify the language and dialect spoken by the user, and quickly retrieve the corresponding text from the language database for output. This seamless switching capability allows users to maintain a smooth interactive experience in different language environments without worrying about language barriers.

[0021] Further, the speech processing module includes a language recognition unit, a dialect recognition unit and a text processing unit;

[0022] A language identification unit, used to identify the user's language type according to the database;

[0023] A dialect recognition unit, used to recognize the user's dialect type according to the database;

[0024] The text processing unit is used to associate the received voice with the database and output the corresponding text after the association is completed.

[0025] The beneficial effect of the basic solution is that by setting up language recognition units and dialect recognition units separately, the system can more accurately identify the language and dialect used by the user. This segmented recognition method reduces recognition errors caused by language or dialect confusion and improves the accuracy and reliability of speech recognition.

[0026] Further, the feature entry module includes a voiceprint feature entry unit, a lip shape feature entry unit, a gesture entry unit, and a private account entry unit;

[0027] A voiceprint feature recording unit, used to identify and record user voiceprint features;

[0028] A mouth shape feature input unit, used to identify and input different mouth shape features of different pronunciations of the user;

[0029] A gesture input unit, used to input user's gesture shortcut commands;

[0030] The private account entry unit is used to enter the user's account name and password and perform identity authentication, and to enter and update the private account's voice input records, text modification records, voice input habits and associated vocabulary library during use.

[0031] The beneficial effect of the basic solution is that through the voiceprint feature recording unit and the lip shape feature recording unit, the system can accurately identify and record the user's unique voiceprint and lip shape features, which provides dual biometric security protection for the user account.

[0032] At the same time, the private account entry unit ensures the privacy of the user's account name and password, as well as the strictness of identity verification, effectively preventing the risk of account theft or misoperation. The lip feature entry unit can assist the voice recognition unit in more accurately identifying the user's pronunciation. By combining voiceprint and lip shape features, the system can more finely distinguish similar pronunciations, reduce misrecognition, and thus improve the accuracy of voice input.

[0033] The gesture input unit allows users to define personalized gesture shortcuts, which greatly simplifies the voice input process. Users do not need to use full voice input every time to convey instructions or information, but can simply make gestures to complete the operation, which improves the convenience and efficiency of interaction.

[0034] The private account entry unit enables the system to have a deeper understanding of the user's usage habits and preferences, thereby providing more personalized and accurate services.

[0035] Furthermore, identity authentication methods include verification code verification, voiceprint verification and face recognition.

[0036] The beneficial effect of the basic solution is that by combining verification code verification, voiceprint verification and face recognition, the system provides a multi-authentication mechanism. This mechanism greatly improves the security of the account. Even if one of the verification methods is compromised, other verification methods can still protect the security of the account. This multiple protection can effectively prevent the risk of unauthorized access and account theft. At the same time, the choice of multiple authentication methods allows users to choose the most suitable verification method according to their preferences and convenience. This design also improves the accuracy and efficiency of verification.

[0037] Furthermore, the account name is composed of the user's primary language, the password is composed of Arabic numerals, and the verification code is composed of English.

[0038] The beneficial effect of the basic solution is: when the user uses it for the first time, a private account needs to be established. At this time, the system can use this environment to know the user's main language, and can separately record the voiceprint characteristics of the user when the voice inputs the main language, when the voice inputs Arabic numerals, and when the voice inputs English. This allows the system to have a preliminary understanding of the user and also facilitates the user's subsequent use of the system.

[0039] Furthermore, gesture shortcut commands include shortcut word input commands, bracelet switch commands, and audio adjustment commands.

[0040] The beneficial effect of the basic solution is that the introduction of gesture shortcuts greatly improves the interaction efficiency between users and the system. Users do not need to use cumbersome voice input or touch screen operations, but can simply make gestures to complete the input of shortcut words, the switch control of the bracelet, and the adjustment of audio, which greatly saves time and energy.

[0041] Further, the correction module includes a speech correction unit and a manual correction unit;

[0042] A voice correction unit, used to receive a voice correction instruction issued by a user for the output text, and automatically correct the text;

[0043] The manual correction unit is used for users to manually modify the output text.

[0044] The beneficial effect of the basic solution is that users can make timely and accurate modifications to the output text, whether through the voice correction unit or the manual correction unit. The voice correction unit can respond to the user's voice modification instructions and automatically adjust the text, reducing the tediousness of manual input by the user. The manual correction unit provides users with a more direct and flexible modification method to ensure the accuracy of the text content.

[0045] Furthermore, the deep learning module includes a context learning unit, a word order learning unit, and an intelligent recommendation unit;

[0046] A context learning unit is used to analyze the interactive context according to the interactive content, extract context keywords according to the context, transmit the context keywords to the text processing unit, and learn the context analysis and context keyword extraction capabilities according to the correctness of the text output;

[0047] The word order learning unit is used to analyze the word order habits of users, analyze and extract word order keywords according to the word order habits of users, transmit the word order keywords to the context learning unit for comparison, and learn the word order habit analysis and word order keyword extraction capabilities according to the comparison similarity;

[0048] The intelligent recommendation unit is used to recommend intelligent keywords to users based on the extracted keywords combined with the associated vocabulary library, and update the associated vocabulary library according to the user's intelligent keyword selection.

[0049] The beneficial effects of the basic solution are: the context learning unit accurately grasps the context by analyzing the interactive content and extracts keywords accordingly. This process greatly enhances the intelligent level of text processing. It can more accurately understand the user's intentions and needs, thereby outputting text content that better meets the user's expectations. At the same time, the context learning unit also has the ability to self-learn, and can continuously optimize its own context analysis and keyword extraction capabilities based on the correctness of the text output, achieving continuous improvement.

[0050] The word order learning unit can gain a deeper understanding of the user's language habits and expressions by analyzing the user's word order habits. This personalized learning enables the system to more accurately understand the user's instructions and needs, and to respond correctly even when faced with non-standard or colloquial input. In addition, the word order learning unit also works in collaboration with the context learning unit to further verify and optimize the analysis results by comparing the similarity of keywords, thereby providing more accurate and personalized services.

[0051] The intelligent recommendation unit recommends related words to users based on the extracted keywords and related vocabulary. This function can not only help users complete input and expression more quickly, but also improve the consistency and fluency of the user experience. Users do not need to search for suitable words or phrases in their minds, the system can automatically provide relevant suggestions, making the entire interaction process smoother and more natural.

[0052] The introduction of deep learning modules enables the system to have the ability of self-learning and optimization. Whether it is the context learning unit, word order learning unit or intelligent recommendation unit, it can continuously adjust and optimize its own algorithms and models based on user feedback and actual conditions. This self-evolution ability enables the system to continuously adapt to new environments and needs, maintaining its competitiveness and advancement.

[0053] Furthermore, a sound receiver and a speaker for voice interaction are fixedly connected inside the bracelet, and a touch screen for text processing is fixedly connected on the surface of the bracelet.

[0054] The beneficial effects of the basic solution are: the design of the bracelet makes the system more convenient to use. The sound receiver and speaker are integrated into the bracelet, so that users can easily receive and play sounds while wearing the bracelet, without having to carry other audio devices, further improving the convenience of the system. In addition, the addition of a touch screen can meet the needs of text output and display, further improving the versatility of the bracelet. Users can perform various operations such as manual correction and gesture instructions by touching the screen. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 This is a flow chart of a system capable of voice input of text in an embodiment of the present invention;

[0056] Figure 2 This is an axonometric diagram of a wristband in an embodiment of the present invention;

[0057] Figure 3 This is a schematic diagram of a speech processing module in an embodiment of the present invention;

[0058] Figure 4 This is a schematic diagram of a feature entry module in an embodiment of the present invention;

[0059] Figure 5 A schematic diagram of a correction module in an embodiment of the present invention;

[0060] Figure 6 Schematic diagram of a deep learning module in an embodiment of the present invention.

[0061] The reference numerals in the drawings of the specification include: wristband 1, touch screen 2. DETAILED DESCRIPTION

[0062] The following is further described in detail through specific implementation methods:

[0063] Embodiment 1:

[0064] Basically as attached Figure 1 and Figure 2 As shown:

[0065] A text input system capable of voice input comprises a bracelet 1, a voice processing module, a feature entry module, a correction module and a deep learning module, wherein the feature entry module, the correction module and the deep learning module are all installed in the bracelet 1, a sound receiver and a speaker for voice interaction are fixedly connected to the bracelet 1 by bolts, and a touch screen 2 for text processing is embedded on the surface of the bracelet 1.

[0066] The specific experimental process is as follows: the user wears the bracelet 1 on the wrist. When the user performs voice interaction for the first time, the voice processing module receives the language spoken by the user, identifies the language and dialect of the language spoken by the user, and compares the language with the language database, retrieves the corresponding text, displays the text, and outputs the output text in the same language and dialect; the user can also choose to convert the text into text in other languages ​​through the touch screen 2 to meet the user's personalized needs.

[0067] During the initial interaction, the voice processing module will guide the user to create a personalized account name and set a private account password. After the setting is completed, the voice processing module will guide the user to enter the specified verification code through voice. During this process, the feature entry module records the user's voiceprint features, lip shape features and facial features. After the feature entry is completed, a private user account is formed in the system. The user can enter the shortcut gesture commands he needs through the gesture entry unit. For example, gesture one represents the user's name. When the user makes gesture one on the bracelet 1, the voice processing module outputs the user's name.

[0068] During daily use, the voice processing module estimates the user's voice input each time to form a voice input record. If the user's voice input is wrong when the text is output, the user can adjust the output text through the correction module, replace the wrong text, sentence or punctuation, or adjust the text format; the correction module will form a correction record based on the user's adjustment.

[0069] And each time the interaction occurs, the system will identify different users based on voiceprint features, and retrieve the corresponding private accounts and personalized settings, thereby achieving the system's targetedness and privacy.

[0070] In daily interactions, the deep learning module will receive the correction records transmitted by the correction module and the voice input records of the voice processing module, analyze the user's commonly used keywords and sentences, and analyze the user's pronunciation in combination with the user's mouth changes, and compare the user's pronunciation with the output text, thereby assisting the voice processing module to output text more accurately; at the same time, the deep learning module will judge the user's emotions and context of speech based on changes in the user's facial expressions, further improving the accuracy of text and punctuation during text output.

[0071] Embodiment 2:

[0072] Basically as attached Figure 3 As shown:

[0073] Different from the above embodiment, the speech processing module is used for speech interaction and text processing and forms a speech input record; the speech processing module is recorded with a language database; the language database includes English, Chinese, French, Russian, Spanish and Arabic dialects and texts. The speech processing module includes a language recognition unit, a dialect recognition unit and a text processing unit.

[0074] The language identification unit is used to identify the user's language type according to the language database.

[0075] The dialect recognition unit is used to recognize the user's dialect type according to the language database.

[0076] The text processing unit is used to associate the received voice with the language database and output the corresponding text after the association is completed.

[0077] The specific experimental process is as follows: When the user interacts with the system through voice, the language recognition unit and the dialect recognition unit can accurately identify the language and dialect used by the user. After the input voice is refined by the language recognition unit and the dialect recognition unit, the text processing unit quickly associates the received voice with the language database according to the search scope of the input voice in the language database, and outputs the associated text. This segmented recognition method reduces the search scope, making the search more efficient, while also reducing recognition errors caused by language or dialect confusion, and improving the accuracy and reliability of voice recognition.

[0078] The present invention has the ability to support globalization by inputting multiple major international languages ​​and their dialects. This allows users to achieve efficient voice input and interaction through the present invention no matter where they are or what language they use, greatly broadening the scope of application and user groups of the system. The system can intelligently identify the language and dialect described by the user, and quickly retrieve the corresponding text from the language database for output. For example, when a user arrives in Sichuan from Beijing, the interactive language changes from Beijing dialect to Sichuan dialect. The dialect recognition unit can quickly identify the change in the input voice and adjust the output voice accordingly. This seamless switching capability allows users to maintain a smooth interactive experience in different language environments without worrying about language barriers.

[0079] Embodiment 3:

[0080] Basically as attached Figure 4 As shown:

[0081] Different from the above-mentioned embodiments, the feature entry module is used to record voiceprint features, lip shape features, facial features and gesture commands, and form a private account based on the recorded voiceprint features, lip shape features, facial features and gesture commands; the feature entry module includes a voiceprint feature entry unit, a lip shape feature entry unit, a gesture entry unit and a private account entry unit.

[0082] The voiceprint feature recording unit is used to identify and record the user's voiceprint features.

[0083] The mouth shape feature input unit is used to identify and input different mouth shape features of different pronunciations of users.

[0084] The gesture input unit is used to input the user's gesture shortcut commands.

[0085] The private account entry unit is used to enter the user's account name and password and perform identity authentication, and to enter and update the private account's voice input records, text modification records, voice input habits and associated vocabulary during use. Identity authentication methods include verification code verification, voiceprint verification and face recognition. The account name is composed of the user's main language, the password is composed of Arabic numerals, and the verification code is composed of English.

[0086] The specific experimental process is as follows: When the user interacts with the system for the first time, the private account entry unit will guide the user to establish a private account. At this time, the user creates an account name in the main language, creates a password in Arabic numerals, and enters a verification code in English. The system can respectively enter the different voiceprint features and lip shape features of the user when using the main language, Arabic numerals, and English. At the same time, the lip shape feature entry unit enters the user's facial features to gain a more comprehensive understanding of the user.

[0087] When a user interacts with the system, the system can accurately identify and record the user's unique voiceprint and lip shape features through the voiceprint feature recording unit and the lip shape feature recording unit, which provides dual biometric security protection for the user account.

[0088] For example, every time user A inputs "yes", the system analyzes that user A pronounces "ah" as one tone, and his mouth shape is relatively flat, with a small degree of lip opening; every time user B inputs "yes", the system analyzes that user B pronounces "ah" as three tones, and his mouth shape is slightly round, with a large degree of lip opening; the system uses this to record different voiceprint features, mouth shape features, and voice input habits for different users.

[0089] At the same time, the private account entry unit ensures the privacy of the user's account name and password, as well as the strictness of identity verification, effectively preventing the risk of account theft or misoperation. The lip feature entry unit not only enhances the security of the account, but also assists the voice recognition unit in more accurately identifying the user's pronunciation. By combining voiceprint and lip shape features, the system can more finely distinguish similar pronunciations, reduce misrecognition, and thus improve the accuracy of voice input.

[0090] The gesture input unit allows users to define personalized gesture shortcuts. For example, if a user enters gesture 1 as turning off the volume in the gesture input unit, when the user is in a meeting, the user only needs to make gesture 1 on bracelet 1, and bracelet 1 automatically turns off the volume; if a user enters gesture 2 as entering the user's address in the gesture input unit, the user only needs to make gesture 2 on bracelet 1, and the text processing unit automatically enters the user's address. This greatly simplifies the voice input process. Users do not need to convey instructions or information through complete voice input every time, they can simply make gestures to complete the operation, which improves the convenience and efficiency of interaction.

[0091] The private account entry unit is not only responsible for entering the user's account name and password, but also for continuously entering and updating the private account's voice input records, text modification records, voice input habits, and associated vocabulary during use. For example, when a user frequently enters words such as hospital, doctor, and surgery, the system will analyze that the user is a medical worker-related group, and will associate vocabulary related to medical workers for the user to use. This comprehensive account management function enables the system to have a deeper understanding of the user's usage habits and preferences, thereby providing more personalized and accurate services.

[0092] For example, users often input sentences such as "Liver: normal size and shape, smooth capsule echo, homogeneous liver parenchyma echo, clear duct structure and normal course, an echo-free area of ​​about 3.2cm×2.5cm in size was detected in the S5 segment of the liver, with regular shape and clear boundaries, CDFI: no obvious blood flow signals were detected in the periphery and inside. The inner diameters of the portal vein and the hepatic vein were not widened." The system will analyze that the user is a doctor, enter the words and sentences commonly used by various doctors into its associated vocabulary library, and form voice input habits based on the word order and frequency of adjective use during input, and provide sentence references for users to choose from in subsequent use.

[0093] Embodiment 4:

[0094] Basically as attached Figure 5 As shown:

[0095] What is different from the above-mentioned embodiment is that the correction module is used to correct the text and form a correction record; the correction module includes a voice correction unit and a manual correction unit.

[0096] The voice correction unit is used to receive the voice correction instruction issued by the user to the output text and automatically correct the text.

[0097] The manual correction unit is used for users to manually modify the output text.

[0098] The specific experimental process is as follows:

[0099] When the user interacts with the system by voice and the text output by the system is wrong, the user can issue a voice command through the voice correction unit to modify the wrong text. For example, if the user issues a modification command "change seal to bee", the correction module will change the seal in the text to bee; if the user issues a modification command "delete bee", the correction module will delete all bees in the text; or if the user issues a modification command "delete the third word in the second line", the correction module will delete the third word in the second line in the text. If the user does not know how to describe the voice modification or thinks that the voice modification is incomplete, the user can also modify the text directly through the manual correction unit.

[0100] Whether through the voice correction unit or the manual correction unit, users can make timely and accurate modifications to the output text. The voice correction unit can respond to the user's voice modification instructions and automatically adjust the text, reducing the tediousness of manual input by the user. The manual correction unit provides users with a more direct and flexible modification method to ensure the accuracy of the text content.

[0101] Embodiment 5:

[0102] Basically as attached Figure 6 As shown:

[0103] Different from the above-mentioned embodiment, the deep learning module includes a context learning unit, a word order learning unit and an intelligent recommendation unit.

[0104] The context learning unit is used to analyze the interactive context according to the interactive content, extract context keywords according to the context, transmit the context keywords to the text processing unit, and learn the context analysis and context keyword extraction capabilities according to the correctness of the text output.

[0105] The word order learning unit is used to analyze the word order habits of users, analyze and extract word order keywords according to the word order habits of users, transmit the word order keywords to the context learning unit for comparison, and learn the word order habit analysis and word order keyword extraction capabilities according to the comparison similarity;

[0106] The intelligent recommendation unit is used to recommend intelligent associated words to users based on the extracted keywords combined with the associated vocabulary library, and update the associated vocabulary library according to the user's intelligent keyword selection.

[0107] The specific experimental process is as follows:

[0108] When the user interacts with the system by voice, the context learning unit accurately grasps the context by analyzing the interactive content and extracts keywords accordingly. This process greatly enhances the intelligent level of text processing. It can more accurately understand the user's intentions and needs, thereby outputting text content that better meets the user's expectations. At the same time, the context learning unit also has self-learning capabilities, and can continuously optimize its own context analysis and keyword extraction capabilities based on the correctness of the text output to achieve continuous improvement. When the user repeatedly mentions "a certain place", "a certain event" or "a certain person" during voice interaction with the system, the context learning unit will extract these keywords and send them to the text processing unit to improve the output accuracy of the text processing unit. If the user modifies the output content, the context learning unit will optimize the learning based on the user's modified content.

[0109] For example, the user inputs "Both kidneys: normal size and shape, clear contours, smooth and regular capsules, slightly high echoes of about 4.2cm×4.5cm in the left renal parenchyma, regular shape, clear boundaries, CDFI: no obvious blood flow signals were detected around and inside the slightly high echoes. The echoes of the remaining parenchyma are homogeneous, the boundary between the cortex and medulla is clear, and the collecting system is not separated. CDFl: No obvious abnormal blood flow signals were found in both kidneys." The system extracts the recurring key sentences "kidney", "echo", and "blood flow signal", and then analyzes and summarizes the sentences, and outputs the key sentence "The left renal parenchyma is slightly high echo." At the same time, the context learning unit optimizes learning based on the key sentences, analyzes that the user should need to judge the kidney condition, and provides the user with relevant conclusions based on the input sentence: "Considered to be a hamartoma."

[0110] In the process of daily voice interaction between users and the system, the word order learning unit can analyze the user's word order habits to gain a deeper understanding of the user's language habits and expressions. The word order learning unit will analyze and judge the input voice based on the user's frequently input voice as a word order template, and when outputting voice and text, the word order learning unit will also output according to the user's word order template. This personalized learning enables the system to understand the user's instructions and needs more accurately, and can respond correctly even when faced with non-standard or colloquial input. In addition, the word order learning unit also works in collaboration with the context learning unit to further verify and optimize the analysis results by comparing the similarity of keywords, thereby providing more accurate and personalized services. Different users have different word orders. The word order learning unit can accurately judge the user's sentence order, thereby extracting the subject, predicate, and object, and quickly extracting the key words in the sentence; for example, some users are accustomed to using inverted word order, and the word order learning unit recognizes that the sentence pattern is subject-postponed, rather than an incorrect sentence pattern. This design improves the recognition rate and inclusiveness of the system.

[0111] Some user-entered sentences have specific formats, and the word order learning unit will also optimize learning. The word order learning unit will extract the same or similar word positions based on the user's multiple inputs to form a word order template. For example, users often enter "Pancreas: normal size and shape, clear boundaries, homogeneous echo, no obvious space inside. The main pancreatic duct is not dilated. CDFI: No obvious abnormal blood flow signal is found in the pancreas." The word order learning unit analyzes the sentence to form a template, "Organ: size and shape analysis, contour analysis, echo analysis, internal structure analysis. Blood vessel analysis. CDFI: blood flow signal analysis." When the user enters other organs again, such as spleen, the sentence "Spleen: normal size and shape, clear boundaries, homogeneous echo of the substance, no obvious space inside. The splenic portal vein is not dilated. CDFI: No obvious abnormal blood flow signal is found in the spleen." will be automatically formed.

[0112] The user can also directly enter the template. For example, the user enters "Pancreas: normal size and shape, clear boundaries, homogeneous echo, and no obvious mass inside. The main pancreatic duct is not dilated. CDFI: No obvious abnormal blood flow signals are found in the pancreas." as a template. The user only needs to enter "pancreas" and the touch screen 2 will automatically display the template. The user only needs to confirm by voice or manually to quickly enter the template, and then quickly modify the input content by modifying the voice modification module or the manual modification module, and the word order learning unit will automatically determine the input position according to the input content.

[0113] Whenever the user interacts with the system by voice, the intelligent recommendation unit recommends related words to the user based on the extracted keywords and related vocabulary. At the same time, the intelligent recommendation unit uses the user's input habits as a template to supplement the incomplete sentences input by the user. This function can not only help users complete input and expression more quickly, but also improve the consistency and fluency of the user experience. Users do not need to search for suitable words or phrases in their minds, the system can automatically provide relevant suggestions, making the entire interaction process smoother and more natural.

[0114] For example, users often input "An echoless liver S5 segment with a size of about 3.2cm×2.5cm, regular shape, clear boundaries, CDFI: no obvious blood flow signal was detected in the periphery and inside." The intelligent recommendation unit will use this sentence as a template; when the user only inputs "S5 segment liver cyst, size 3.2×2.5", the intelligent recommendation unit will automatically extract the main keyword "liver", the location keyword "S5 segment", the status keyword "cyst" and "size 3.2×2.5", and then automatically complete the sentence according to the user's usual input habits and sentence format, and display "liver" on the touch screen 2. S5 segment detected an echoless object of about 3.2cm×2.5cm in size, with regular shape and clear boundaries. CDFI: no obvious blood flow signal was detected in the periphery and inside. ", and the user can customize the input content and input the specified position through the voice correction unit or the manual correction unit. After the input is completed, the user can confirm by voice or manually to complete the input of the recommended sentence. After the input is completed, the system will also analyze the relevant results based on the input content. For example, after the user enters "gallbladder stone diameter 2.5", the touch screen will display "a strong echo of about 2.5cm in diameter was detected in the gallbladder cavity, accompanied by an acoustic shadow behind.".

[0115] In subsequent use, the user only needs to modify the input part and part size. The system will automatically associate the voice templates that the user frequently inputs based on the input part and part size, replace only the part and part size, output the corresponding sentence, and associate the corresponding conclusion based on the sentence.

[0116] The intelligent recommendation unit will also recommend multiple recommended sentences for users to choose according to the input frequency. For example, if "with echo" and "without echo" are frequently input, and the frequency of inputting "with echo" is greater than the frequency of inputting "without echo", the intelligent recommendation unit will automatically recommend 1. "with echo" and 2. "without echo" on the touch screen according to the input frequency. Users can select by voice input number or manually click to select.

[0117] The introduction of deep learning modules enables the system to have the ability of self-learning and optimization. Whether it is the context learning unit, word order learning unit or intelligent recommendation unit, it can continuously adjust and optimize its own algorithms and models based on user feedback and actual conditions. This self-evolution ability enables the system to continuously adapt to new environments and needs, maintaining its competitiveness and advancement.

[0118] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0119] The above is only an embodiment of the present invention. The common sense such as the known specific structure and characteristics in the scheme is not described in detail here. The ordinary technicians in the relevant field know all the common technical knowledge in the technical field of the invention before the application date or priority date, can obtain all the existing technologies in the field, and have the ability to apply the conventional experimental means before that date. The ordinary technicians in the relevant field can improve and implement this scheme in combination with their own abilities under the enlightenment given by this application. Some typical known structures or known methods should not become obstacles for ordinary technicians in the relevant field to implement this application. It should be pointed out that for those skilled in the art, without departing from the structure of the present invention, several deformations and improvements can be made, which should also be regarded as the protection scope of the present invention, which will not affect the effect of the implementation of the present invention and the practicality of the patent. The protection scope required by this application shall be based on the content of its claims, and the specific implementation methods and other records in the specification can be used to interpret the content of the claims.

Claims

1. A system capable of voice inputting text, characterized in that: It includes a bracelet, a voice processing module, a feature entry module, a correction module and a deep learning module; The feature entry module, correction module and deep learning module are all installed in the bracelet; A speech processing module is used for speech interaction and text processing, and forms a speech input record; the speech processing module is input with a language database; A feature input module, used to input voiceprint features, lip shape features, facial features and gesture commands, and to form a private account based on the input voiceprint features, lip shape features, facial features and gesture commands; A correction module is used to correct the text and form a correction record; The deep learning module is used to optimize the system according to the user's voice input habits, interaction context and speaking style, and analyze the voice input records and correction records of private accounts to form an input habit library and associated vocabulary library of private accounts; and transmit the relevant input habits and associated vocabulary to the voice processing module during the text processing process.

2. The system for voice input of text according to claim 1, characterized in that: The language database includes dialects and scripts of English, Chinese, French, Russian, Spanish and Arabic.

3. The system for voice input of text according to claim 2, characterized in that: The speech processing module includes a language recognition unit, a dialect recognition unit and a text processing unit; A language identification unit, used to identify the user's language type based on the language database; A dialect identification unit, used to identify the user's dialect type according to the language database; The text processing unit is used to associate the received voice with the language database and output the corresponding text after the association is completed.

4. The system for voice input of text according to claim 3, characterized in that: The feature input module includes a voiceprint feature input unit, a lip shape feature input unit, a gesture input unit, and a private account input unit; A voiceprint feature recording unit, used to identify and record user voiceprint features; A mouth shape feature input unit, used to identify and input different mouth shape features of different pronunciations of the user; A gesture input unit, used to input user's gesture shortcut commands; The private account entry unit is used to enter the user's account name and password and perform identity authentication, and to enter and update the private account's voice input records, text modification records, voice input habits and associated vocabulary library during use.

5. The system for voice input of text according to claim 4, characterized in that: Authentication methods include verification code verification, voiceprint verification and face recognition.

6. The system for voice input of text according to claim 5, characterized in that: The account name is composed of the user's primary language, the password is composed of Arabic numerals, and the verification code is composed of English.

7. A system for voice input of text according to claim 6, characterized in that: Gesture shortcut commands include quick phrase input commands, bracelet switch commands, and audio adjustment commands.

8. A system for voice input of text according to claim 7, characterized in that: The correction module includes a speech correction unit and a manual correction unit; A voice correction unit, used to receive a voice correction instruction issued by a user for the output text, and automatically correct the text; The manual correction unit is used for users to manually modify the output text.

9. A system for voice input of text according to claim 8, characterized in that: The deep learning module includes context learning unit, word order learning unit and intelligent recommendation unit; A context learning unit is used to analyze the interaction context according to the interaction content, extract context keywords according to the context, transmit the context keywords to the text processing unit, and learn the context analysis and context keyword extraction capabilities according to the text output; The word order learning unit is used to analyze the word order habits of users, analyze and extract word order keywords according to the word order habits of users, transmit the word order keywords to the context learning unit for comparison, and learn the word order habit analysis and word order keyword extraction capabilities according to the comparison similarity; The intelligent recommendation unit is used to recommend intelligent keywords to users based on the extracted keywords combined with the associated vocabulary library, and update the associated vocabulary library according to the user's intelligent keyword selection.

10. A system for voice input of text according to claim 9, characterized in that: A sound receiver and a speaker for voice interaction are fixedly connected inside the bracelet, and a touch screen for text processing is fixedly connected on the surface of the bracelet.

Images