A human-computer interaction system based on speech recognition
By collecting and analyzing user voice information in the speech recognition system, combining forward-tracking voice information and functional link list, the problem of insufficient recognition accuracy in the prior art is solved, and higher speech information recognition accuracy and auxiliary display are achieved Recommended accuracy.
Patent Information
- Application Number
- CN202411150352.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-21
- Publication Date
- 2024-11-15
- Estimated Expiration
- 2044-08-21
AI Technical Summary
The existing human-computer interaction system based on speech recognition has the problem of poor recognition accuracy when users apply for functions. Especially when similar functional texts exist, they fail to effectively consider the relevance and repetition of adjacent instructions, resulting in The recognition accuracy of the recognition function is relatively low.
By introducing a voice acquisition module, recognition matching unit and interactive display unit in the voice recognition system, users' voice information is collected and recognized, and functions are confirmed and auxiliary display selection are performed in combination with forward-tracking voice information, and cascading function tables and recognition functions are adopted. The linked list performs functional matching and analysis, considers the similarity and correlation of voice information, optimizes functional recognition and assists display.
Improve the accuracy and accuracy of voice information recognition function, enhance the analysis of user function use relevance and repetition, provide more accurate and convenient auxiliary display of selected functions, and improve user's next voice information Identify accuracy.
Smart Images

Figure CN118658479B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of speech recognition, and in particular to a human-computer interaction system based on speech recognition. Background Art
[0002] As information technology develops rapidly and the intelligent era has arrived, human-computer interaction modes continue to evolve and innovate. Traditional human-computer interaction methods, such as keyboard input and mouse clicks, have gradually revealed their limitations in some situations and are unable to meet people's desire for more natural, convenient and efficient forms of interaction.
[0003] In this situation, speech recognition technology has gradually become the research focus in the field of human-computer interaction. As the most natural and direct way of communication for humans, speech has the outstanding advantages of rapid and convenient information transmission without additional learning costs. The human-computer interaction system based on speech recognition aims to achieve more natural, smooth and intelligent communication between humans and machines by understanding and processing human speech.
[0004] However, an existing human-computer interaction system based on voice recognition has obvious defects when a user applies to use a function. It collects the user's voice information, uses voice recognition technology to recognize it, and then compares the similarity with the text information of several functions pre-stored internally, and selects the function corresponding to the text information with the highest similarity as the recognition function for the user to recognize, confirm and finally execute. However, when there are several similar function text information, the accuracy of this method is poor, and this method does not consider the correlation of adjacent instructions. At the same time, the use of functions by authorized users is usually relevant and repetitive, but the current technology does not apply this to the selection of recognition functions, which undoubtedly leads to a low recognition accuracy rate of the recognition function.
[0005] In order to solve the above problems, the present invention proposes a solution. Summary of the invention
[0006] The purpose of the present invention is to provide a human-computer interaction system based on speech recognition, in order to solve the problems raised in the above background technology;
[0007] The purpose of the present invention can be achieved through the following technical solutions:
[0008] A human-computer interaction system based on speech recognition, comprising:
[0009] Voice collection module, used to collect voice information sent by authorized users;
[0010] The recognition and matching unit is used to recognize and obtain the characteristic text of the current voice information after receiving the voice information sent by the authorized user at the current moment;
[0011] According to the time when the characteristic text of the current voice information is obtained, a number of voice information sent by the authorized user within the forward P1 period is obtained, and the number of voice information sent by the authorized user within the forward P1 period is determined to be non-zero, and a function to be confirmed for obtaining the current voice information is selected based on the determination result;
[0012] The interactive display unit displays the to-be-confirmed function of the current voice information to the authorized user for function confirmation. After the authorized user completes the confirmation, the to-be-confirmed function of the current voice information is used as the recognition function of the current voice information;
[0013] The function realization unit executes the corresponding realization logic according to the recognition function of the current voice information. During the execution, the interactive display unit obtains a number of voice messages sent by the authorized user within the forward P3 time according to the sending time of the current voice information. P3 is a preset auxiliary display selected reference time threshold;
[0014] A non-zero determination is made on the number of voice messages sent by the authorized user within the forward P3 time period, and based on the determination result, three auxiliary display selection functions are selected for the next voice message sent by the authorized user, and their feature texts are displayed to the authorized user for viewing, so as to provide assistance for the next voice message sent by the authorized user.
[0015] Further, if the number of voice messages sent by the authorized user within the time period P1 is not 0, the steps of selecting the function of obtaining the pending confirmation of the current voice message are as follows:
[0016] S21: The recognition matching unit obtains the recognition function of a voice message that is closest in time to the current voice message from all voice messages sent by all authorized users within the period P1, and uses the recognition function as the identification selection function;
[0017] S22: According to the cascading function table identifying the selected function, the feature texts of all the cascading functions included in the cascading function table are extracted, and the similarities between the feature texts of all the cascading functions included in the cascading function table and the feature text of the current voice information are calculated in sequence from top to bottom according to the arrangement order of each cascading function in the cascading function table. Each time a similarity is calculated, the similarity is compared with P2. If the similarity is greater than P2, the cascading function corresponding to the feature text participating in the similarity calculation is used as the function to be confirmed of the current voice information, and the similarity calculation between the feature texts of the remaining cascading functions and the feature text of the current voice information is stopped. P2 is a preset selected similarity comparison threshold.
[0018] Further, in S22, if the similarities between the feature texts of all cascade functions included in the cascade function table and the feature text of the current voice information are less than or equal to P2 after the similarity calculation is completed, then the feature texts of all business functions are extracted from top to bottom in sequence according to the arrangement order of all business functions in the function frequency table, and each time the feature text of a business function is extracted, the similarity between the feature text and the feature text of the current voice information is calculated, and after the feature texts of all business functions are completely extracted, the business function corresponding to the feature text participating in the similarity calculation with the largest value is obtained as the function to be confirmed for the current voice information.
[0019] Further, if the number of voice messages sent by the authorized user within the time P3 is not 0, the steps of selecting three auxiliary display selection functions for the next voice message sent by the authorized user are as follows:
[0020] S31: according to the time sequence of voice messages being sent, the recognition functions of all voice messages sent by the authorized user within the period P3 are marked as H1, H2, ..., Hh, where h≥1;
[0021] S32: Obtaining from the interactive display unit all recognition function lists containing the recognition function of the current voice information, marked as I1, I2, ..., Ii, i≥1 respectively;
[0022] S33: Calculate and obtain the difference indexes L1, L2, ..., Lh of the recognition functions H1, H2, ..., Hh based on the recognition function chain list I1 according to the preset calculation rules. The calculation rules are as follows:
[0023] S331: extracting all recognition functions arranged on the left side of the recognition function of the current voice information from the recognition function chain list I1 in order from left to right, and marking all the extracted recognition functions as J1, J2, ..., Jj in order of extraction, where j≥1;
[0024] S332: sequentially compare the identification functions J1, J2, ..., Jj with the identification function H1 for consistency, and obtain the first identification function subscript K1 that is consistent with the identification function H1;
[0025] S333: Calculate and obtain the difference index L1 of the recognition function H1 based on the recognition function linked list I1 using the formula L1=|K1-(h-1)|. If there is no recognition function consistent with the recognition function H1 after traversing all the recognition functions J1, J2, ..., Jj, then assign P4 to the difference index L1 of the recognition function H1 based on the recognition function linked list I1, where P4 is a preset difference equalization index.
[0026] S334: according to S331 to S333, the difference indexes L2, L3, ..., Lh of the recognition functions H2, H3, ..., Hh based on the recognition function linked list I1 are calculated in sequence;
[0027] S34: Utilize the formula Calculate and obtain the auxiliary evaluation index N1 of the current voice information based on the recognition function list I1, where M1 in the formula is the total number of recognition functions in the recognition functions J1, J2, ..., Jj that are inconsistent with any of the recognition functions H1, H2, ..., Hh, ɑ1 is the preset harmony compensation coefficient, Ll refers to the difference index L1, L2, ..., Ll, and the value of l is 1, 2, ..., h;
[0028] S35: according to S31 to S34, the auxiliary evaluation indexes N2, N3, ..., Ni of the current voice information based on the recognition function chain lists I2, I3, ..., Ii are calculated in sequence, and the recognition function chain lists corresponding to the auxiliary evaluation indexes with the first smallest, the second smallest, and the third smallest values are selected as the auxiliary chain lists for the next voice information sent by the authorized user;
[0029] For an auxiliary linked list selected as the next voice message sent by the authorized user, the recognition function located to the right of the first current voice information recognition function is extracted from left to right as an auxiliary display selected function for the next voice message sent by the authorized user. Similarly, three auxiliary display selected functions for the next voice message sent by the authorized user can be obtained.
[0030] Beneficial effects of the present invention:
[0031] (1) The present invention collects voice information sent by the authorized user by setting a voice collection module, the recognition and matching unit recognizes the voice information and displays it on the interactive display unit for the authorized user to confirm the recognition function, the function realization unit realizes the recognition function logically, and the interactive display unit selects an auxiliary display function for the next voice information sent by the authorized user based on the recognition function of the current voice information. In this way, the authorized user is assisted in sending the next voice information, and the recognition function of the next voice information is optimized;
[0032] (2) In the process of recognizing voice information through the recognition and matching unit of the present invention, the voice recognition technology is used to recognize the feature text of the voice information, and the voice information traced back to the time P1 is obtained based on the time when the current voice information is sent, and the function to be confirmed of the current voice information is determined based on the number of voice information and the corresponding recognition function. In this way, the influence of the recognition function with similar feature text on the recognition result is avoided, the accuracy of the voice information feature text matching recognition function is enhanced, and the accuracy of voice recognition is further improved;
[0033] (3) The present invention deconstructs and analyzes the recognition functions of all voice information sent by the authorized user within a number of voice interaction cycles by setting a function deconstruction module, determines a cascade function table based on all recognition functions of the authorized user and a recognition function chain table of a number of voice interaction cycles, and combines the recognition function of a single voice information sent by the authorized user and the analysis of a single voice information and its adjacent voice information when determining the cascade function table. In this way, the selection of the function to be confirmed of the current voice information is made more accurate and comprehensive, which not only fully considers the independent characteristics of a single voice information, but also takes into account its association and influence with adjacent voice information, thereby effectively improving the accuracy and effectiveness of the voice information recognition function;
[0034] (4) During the execution of the recognition function of the current voice information, the present invention synchronously traces back several voice information within a preset time and combines the recognition function of the current voice information in the recognition function chain list of several voice interaction cycles to provide auxiliary display selected function recommendations for the authorized user to send out the voice information in sequence next time. In this way, by combining the recognition function corresponding to the authorized user's previous voice information, the auxiliary display selected function recommendation for the authorized user is integrated with the authorized user's analysis of the relevance and repeatability of the function, which greatly improves the accuracy of the auxiliary display selected function recommendation and also provides convenience for the user to send out the voice message next time. BRIEF DESCRIPTION OF THE DRAWINGS
[0035] The present invention will be further described below in conjunction with the accompanying drawings.
[0036] Figure 1 is a system block diagram of the present invention;
[0037] Figure 2 It is a flow chart of the method of the present invention. DETAILED DESCRIPTION
[0038] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0039] like Figure 1 , 2 As shown, a human-computer interaction system based on speech recognition includes a speech acquisition module, a speech interaction module and a function decomposition module;
[0040] The voice collection module is used to collect voice information of authorized users. In this embodiment, the authorized users refer to users who have passed facial recognition authentication;
[0041] After the authorized user issues a voice service function instruction at the current moment, the voice collection module receives the issued voice service function instruction and collects the voice information sent by the authorized user after receiving it, and the voice collection module transmits the voice information at the current moment to the voice interaction module;
[0042] A voice interaction module, used to identify and match the voice information of the authorized user, the voice interaction module includes an identification and matching unit, an optimization and matching unit, an interactive display unit and a function implementation unit;
[0043] The recognition and matching unit pre-stores a function frequency table, characteristic texts of several business functions, and a cascade function table; in this embodiment, several business functions are provided to meet the specific needs of authorized users;
[0044] After receiving the voice information transmitted at the current moment, the voice interaction module transmits it to the recognition and matching unit. After receiving the voice information transmitted at the current moment, the recognition and matching unit uses voice recognition technology to recognize the voice information to obtain the feature text of the current voice information;
[0045] The recognition and matching unit obtains a number of voice messages sent by the authorized user within a period of time P1 according to the time when the feature text of the current voice message is obtained, where P1 is a preset period of time for retrospective recognition association;
[0046] If the number of voice messages sent by the authorized user within the time period P1 is 0, then the feature texts of all business functions are extracted from top to bottom in sequence according to the arrangement order of all business functions in the function frequency table, and each time a feature text of a business function is extracted, the feature text and the feature text of the current voice message are similarly calculated, and after all the feature texts of the business functions are completely extracted, the business function corresponding to the feature text participating in the similarity calculation with the largest value is obtained as the function to be confirmed of the current voice message and transmitted to the interactive display unit;
[0047] If the number of voice messages sent by the authorized user within the time period P1 is not 0, the pending confirmation function of the current voice message is selected according to the preset matching selection rule. The matching selection rule is as follows:
[0048] S21: The recognition matching unit obtains the recognition function of a voice message that is closest in time to the current voice message from all voice messages sent by all authorized users within the period P1, and uses the recognition function as the identification selection function;
[0049] S22: according to the cascading function table identifying the selected function, extracting the feature texts of all the cascading functions contained in the cascading function table, calculating the similarity between the feature texts of all the cascading functions contained in the cascading function table and the feature text of the current voice information in turn from top to bottom according to the arrangement order of each cascading function in the cascading function table, and comparing the similarity with P2 for each similarity calculated, if the similarity is greater than P2, transmitting the cascading function corresponding to the feature text involved in the similarity calculation as the to-be-confirmed function of the current voice information to the interactive display unit, and stopping the similarity calculation between the feature texts of the remaining cascading functions and the feature text of the current voice information;
[0050] If the similarities between the characteristic texts of all cascade functions included in the cascade function table and the characteristic text of the current voice information are calculated to be less than or equal to P2, P2 is a preset selected similarity comparison threshold;
[0051] Then, according to the arrangement order of all business functions in the function frequency table, the feature texts of all business functions are extracted from top to bottom, and each time the feature text of a business function is extracted, the feature text and the feature text of the current voice information are similarity calculated, and after all the feature texts of the business functions are completely extracted, the business function corresponding to the feature text participating in the similarity calculation with the largest value is obtained as the function to be confirmed of the current voice information and transmitted to the interactive display unit;
[0052] It should be noted that, on the basis that the similarities between the feature texts of all cascade functions in the cascade function table and the feature text of the currently received voice information are less than or equal to P2, in the subsequent process of extracting the similarity with the largest value, the similarities between the feature texts of all cascade functions in the cascade function table and the feature text of the currently received voice information are also involved in the similarity comparison and extraction process;
[0053] The interactive display unit displays the to-be-confirmed function of the current voice information to the authorized user for function confirmation. After the authorized user completes the confirmation, the to-be-confirmed function of the current voice information is transmitted to the function realization unit as the recognition function of the current voice information;
[0054] The function realization unit pre-stores the realization logic of all business functions. After receiving the recognition function of the transmitted current voice information, the function realization unit searches for the corresponding realization logic and executes the realization logic;
[0055] During the execution process, the interactive display unit obtains a number of voice messages sent by the authorized user within a period of P3 according to the sending time of the current voice message, where P3 is a preset reference time threshold for auxiliary display selection;
[0056] If the number of voice messages sent by the authorized user during the period P3 is not 0, the next voice message sent by the authorized user is assisted according to the preset auxiliary display selection rule. The auxiliary display selection rule is as follows:
[0057] S31: according to the time sequence of voice information issuance, the recognition functions of all voice information issued by the authorized user within the period P3 are marked as H1, H2, ..., Hh in sequence, where h is the total number of voice information issued by the authorized user within the period P3 from the time when the current voice information is issued, and h≥1, wherein the voice information with the recognition function H1 is issued first, and the voice information with the recognition functions H2, H3, ..., Hh is issued in the order of H1, H2, ..., Hh;
[0058] S32: Acquire all recognition function chain lists containing the recognition function of the current voice information from the interactive display unit, marked as I1, I2, ..., Ii, i≥1, i is the total number of recognition function chain lists containing the recognition function of the current voice information acquired in the interactive display unit;
[0059] S33: Calculate and obtain the difference indexes L1, L2, ..., Lh of the recognition functions H1, H2, ..., Hh based on the recognition function chain list I1 according to the preset calculation rules. The calculation rules are as follows:
[0060] S331: extracting all recognition functions arranged on the left side of the recognition function of the current voice information from the recognition function chain list I1 in order from left to right, and marking all the extracted recognition functions as J1, J2, ..., Jj in order of extraction, where j≥1, and j is the total number of recognition functions extracted from the recognition function chain list I1 and arranged on the left side of the recognition function of the current voice information;
[0061] S332: Compare the identification functions J1, J2, ..., Jj with the identification function H1 in sequence for consistency, and obtain the tag subscript K1 of the first identification function that is consistent with the identification function H1, wherein consistency refers to the same identification function using different tags, and the content and implementation logic are the same;
[0062] S333: Calculate and obtain the difference index L1 of the recognition function H1 based on the recognition function list I1 using the formula L1=|K1-(h-1)|;
[0063] It should be noted here that if there is no recognition function consistent with the recognition function H1 after all the recognition functions J1, J2, ..., Jj are traversed completely, P4 is assigned to the difference index L1 of the recognition function H1 based on the recognition function linked list I1, and P4 is the preset difference equalization index;
[0064] S334: according to S331 to S333, the difference indexes L2, L3, ..., Lh of the recognition functions H2, H3, ..., Hh based on the recognition function linked list I1 are calculated in sequence;
[0065] S34: Utilize the formula Calculate and obtain the auxiliary evaluation index N1 of the current voice information based on the recognition function list I1, where M1 in the formula is the total number of recognition functions in the recognition functions J1, J2, ..., Jj that are inconsistent with any of the recognition functions H1, H2, ..., Hh, ɑ1 is the preset harmony compensation coefficient, Ll refers to the difference index L1, L2, ..., Ll, and the value of l is 1, 2, ..., h;
[0066] S35: according to S31 to S34, the auxiliary evaluation indexes N2, N3, ..., Ni of the current voice information based on the recognition function chain lists I2, I3, ..., Ii are calculated in sequence, and the recognition function chain lists corresponding to the auxiliary evaluation indexes with the first smallest, the second smallest, and the third smallest values are selected as the auxiliary chain lists for the next voice information sent by the authorized user;
[0067] For an auxiliary chain list selected as the next voice message sent by the authorized user, the recognition function located on the right adjacent to the first current voice message recognition function is extracted from left to right as an auxiliary display selected function for the next voice message sent by the authorized user. Similarly, three auxiliary display selected functions for the next voice message sent by the authorized user can be obtained, and their feature texts are displayed to the authorized user for viewing, so as to provide assistance for the authorized user to send the voice message next time;
[0068] If the number of voice messages sent by the authorized user within the time P3 is 0, the cascade function table of the recognition function of the current voice message is obtained from the recognition and matching unit, and the cascade functions with the first, second and third cascade weights are extracted from the table and selected as the three auxiliary display selected functions for the next voice message sent by the authorized user, and the feature texts thereof are displayed to the authorized user for viewing, so as to provide assistance for the authorized user to send the voice message next time;
[0069] The function deconstruction module periodically deconstructs the recognition functions of several voice messages sent by authorized users. The deconstruction steps are as follows:
[0070] S11: Obtain recognition functions of all voice information sent by authorized users in t voice interaction cycles and deduplicate them, and mark all remaining recognition functions after deduplication as A1, A2, ..., Aa, where a≥1, respectively, and the t voice interaction cycles are from the current voice interaction cycle to the past t voice interaction cycles;
[0071] It should be noted here that in a voice interaction cycle, the time interval between the sending time of all voice messages sent by the authorized user and the sending time of each adjacent voice message does not exceed P4, where P4 is the preset standard cycle identification time threshold;
[0072] S12: Selecting the recognition function A1 as the function to be deconstructed, and selecting a voice interaction cycle as the cycle to be deconstructed;
[0073] Obtaining the sending time of the voice information corresponding to the function to be deconstructed from the period to be deconstructed, and taking the sending time as the standard time, selecting the recognition function corresponding to the voice information whose sending time is closest to the standard time and sent after the standard time from the period to be deconstructed, and using the selected recognition function as the cascade function of the function to be deconstructed based on the period to be deconstructed;
[0074] At the same time, a recognition function chain list of the period to be deconstructed is constructed according to the recognition functions of all voice information sent by the authorized user in the period to be deconstructed, and all recognition functions in the recognition function chain list are sorted from left to right in the order of the distance between the sending time of the corresponding voice information and the current time, that is, the closer to the left in the recognition function chain list, the farther the sending time of the corresponding voice information is from the current time;
[0075] S13: Select t voice interaction cycles as to-be-deconstructed cycles in sequence according to S11 to S12, obtain cascade functions of the to-be-deconstructed functions based on the t voice interaction cycles and deduplicate them, and mark all remaining cascade functions after deduplicate as B1, B2, ..., Bb, where b≥1;
[0076] At the same time, a recognition function list of t voice interaction cycles is constructed;
[0077] S14: using the formula Cc=Bc / t to sequentially calculate and obtain the cascade weights C1, C2, ..., Cb of the cascade functions B1, B2, ..., Bb based on the function to be deconstructed, where Bc in the formula refers to each of the cascade functions B1, B2, ..., Bb, and the value of c is 1, 2, ..., b;
[0078] Generate a cascade function table of the functions to be deconstructed based on the cascade weights of the functions to be deconstructed according to the cascade functions B1, B2, ..., Bb;
[0079] The cascade function table includes a cascade weight field and a cascade function field, the cascade weight field stores cascade weights C1, C2, ..., Cb, the cascade weights stored in the cascade weight field decrease in value from top to bottom in the cascade function table, and the cascade function field stores the corresponding cascade function;
[0080] S15: Obtain the total number D1 of recognition functions to be deconstructed from the recognition functions of all voice messages sent by authorized users in t voice interaction cycles, and calculate the frequency weight E1 of the function to be deconstructed using the formula E1=D1 / D, where D is the total number of all voice messages sent by authorized users in t voice interaction cycles;
[0081] S16: sequentially selecting recognition functions A2, A3, ..., Aa as functions to be deconstructed, and sequentially generating a cascade function table of recognition functions A2, A3, ..., Aa and their frequency weights according to S13 to S15;
[0082] S17: Generate a function frequency table according to the frequency weights of the identification functions A1, A2, ..., Aa, wherein the function frequency table includes a frequency weight field and an identification function field, wherein the frequency weight field stores the frequency weights of the identification functions A1, A2, ..., Aa, wherein the frequency weights stored in the frequency weight field decrease in value from top to bottom in the function frequency table, and the identification function field stores the corresponding identification functions;
[0083] The function deconstruction module transmits the generated function frequency table and the cascade function table of several recognition functions to the recognition matching unit for updating and storage, and transmits the recognition function chain table of t voice interaction cycles to the interaction display unit for updating and storage;
[0084] In the description of the specification, the description with reference to the terms "one embodiment", "example", "specific example", etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.
[0085] The above contents are merely examples and explanations of the present invention. Those skilled in the art may make various modifications or additions to the specific embodiments described or replace them in a similar manner. As long as they do not deviate from the invention or exceed the scope defined by the claims, they shall all fall within the protection scope of the present invention.
[0086] The above is a detailed description of an embodiment of the present invention, but the content is only a preferred embodiment of the present invention and cannot be considered to limit the scope of implementation of the present invention. All equivalent changes and improvements made within the scope of the present invention should still fall within the scope of the patent coverage of the present invention.
Claims
1. A human-computer interaction system based on speech recognition, It is characterized in that include: The voice collection module is used to collect the voice information sent by the authorized user; The recognition and matching unit is used to recognize and obtain the characteristic text of the current voice information after receiving the voice information sent by the authorized user at the current moment; According to the time when the characteristic text of the current voice information is obtained, a number of voice information sent by the authorized user within the forward P1 period is obtained, and the number of voice information sent by the authorized user within the forward P1 period is determined to be non-zero, and a function to be confirmed for obtaining the current voice information is selected based on the determination result; The interactive display unit displays the to-be-confirmed function of the current voice information to the authorized user for function confirmation. After the authorized user completes the confirmation, the to-be-confirmed function of the current voice information is used as the recognition function of the current voice information; The function realization unit executes the corresponding realization logic according to the recognition function of the current voice information. During the execution, the interactive display unit obtains a number of voice messages sent by the authorized user within the forward P3 time according to the sending time of the current voice information. P3 is a preset auxiliary display selected reference time threshold; The number of voice messages sent by the authorized user within the forward P3 period is determined to be non-zero, and three auxiliary display selection functions for the next voice message sent by the authorized user are selected based on the determination result, and the feature text thereof is displayed to the authorized user for viewing; If the number of voice messages sent by the authorized user within the time period P1 is not 0, the steps of selecting the function of obtaining the pending confirmation of the current voice message are as follows: S21: The recognition matching unit obtains the recognition function of a voice message that is closest in time to the current voice message from all voice messages sent by all authorized users within the period P1, and uses the recognition function as the identification selection function; S22: according to the cascading function table identifying the selected function, extract the feature texts of all the cascading functions contained in the cascading function table, and calculate the similarity between the feature texts of all the cascading functions contained in the cascading function table and the feature text of the current voice information in turn from top to bottom according to the arrangement order of each cascading function in the cascading function table, and compare the similarity with P2 every time a similarity is calculated, if the similarity is greater than P2, the cascading function corresponding to the feature text involved in the similarity calculation is used as the function to be confirmed of the current voice information, and the similarity calculation between the feature texts of the remaining cascading functions and the feature text of the current voice information is stopped at the same time, and P2 is the preset selected similarity comparison threshold; If the number of voice messages sent by the authorized user within the time P1 is 0, the steps of selecting the function to be confirmed of the current voice message are as follows: according to the arrangement order of all business functions in the function frequency table, the feature texts of all business functions are extracted from top to bottom, and each time the feature text of a business function is extracted, the similarity between the feature text and the feature text of the current voice message is calculated, and after the feature texts of all business functions are completely extracted, the business function corresponding to the feature text participating in the similarity calculation with the largest value is obtained as the function to be confirmed of the current voice message; If the number of voice messages sent by the authorized user within the P3 period is not 0, the steps of selecting three auxiliary display selection functions for the next voice message sent by the authorized user are as follows: S31: according to the time sequence of voice messages being sent, the recognition functions of all voice messages sent by the authorized user within the period P3 are marked as H1, H2, ..., Hh, where h≥1; S32: Obtaining from the interactive display unit all recognition function lists containing the recognition function of the current voice information, marked as I1, I2, ..., Ii, i≥1 respectively; S33: Calculate and obtain the difference indexes L1, L2, ..., Lh of the recognition functions H1, H2, ..., Hh based on the recognition function linked list I1 according to the preset calculation rules. The calculation rules are as follows: S331: extracting all recognition functions arranged on the left side of the recognition function of the current voice information from the recognition function chain list I1 in order from left to right, and marking all the extracted recognition functions as J1, J2, ..., Jj in order of extraction, where j≥1; S332: sequentially compare the identification functions J1, J2, ..., Jj with the identification function H1 for consistency, and obtain the first identification function subscript K1 that is consistent with the identification function H1; S333: Calculate and obtain the difference index L1 of the recognition function H1 based on the recognition function linked list I1 using the formula L1=|K1-(h-1)|. If there is no recognition function consistent with the recognition function H1 after traversing all the recognition functions J1, J2, ..., Jj, then assign P4 to the difference index L1 of the recognition function H1 based on the recognition function linked list I1, where P4 is a preset difference equalization index. S334: according to S331 to S333, the difference indexes L2, L3, ..., Lh of the recognition functions H2, H3, ..., Hh based on the recognition function linked list I1 are calculated in sequence; S34: Utilize the formula Calculate and obtain the auxiliary evaluation index N1 of the current voice information based on the recognition function list I1, where M1 in the formula is the total number of recognition functions in the recognition functions J1, J2, ..., Jj that are inconsistent with any of the recognition functions H1, H2, ..., Hh, ɑ1 is the preset harmony compensation coefficient, Ll refers to the difference index L1, L2, ..., Ll, and the value of l is 1, 2, ..., h; S35: according to S31 to S34, the auxiliary evaluation indexes N2, N3, ..., Ni of the current voice information based on the recognition function chain lists I2, I3, ..., Ii are calculated in sequence, and the recognition function chain lists corresponding to the auxiliary evaluation indexes with the first smallest, the second smallest, and the third smallest values are selected as the auxiliary chain lists for the next voice information sent by the authorized user; For an auxiliary chain list selected as the next voice message sent by the authorized user, extract the recognition function located on the right adjacent to the first current voice message recognition function in order from left to right as an auxiliary display selected function for the next voice message sent by the authorized user. Similarly, three auxiliary display selected functions for the next voice message sent by the authorized user can be obtained; If the number of voice messages sent by the authorized user within the time P3 is 0, the steps for selecting three auxiliary display selected functions for the next voice message sent by the authorized user are as follows: obtain the cascade function table of the recognition function of the current voice information from the recognition matching unit, and extract the cascade functions with the first, second and third cascade weights therefrom as the three auxiliary display selected functions for the next voice message sent by the authorized user, and display their feature texts to the authorized user for viewing, so as to provide assistance for the authorized user to send a voice message next time.
2. A human-computer interaction system based on speech recognition according to claim 1, It is characterized in that It also includes a function deconstruction module, which performs function deconstruction on the recognition function of several voice messages sent by the authorized user, and the steps are as follows: S11: Obtain recognition functions of all voice information sent by authorized users in t voice interaction cycles and deduplicate them, and mark all remaining recognition functions after deduplication as A1, A2, ..., Aa, where a≥1, respectively, and the t voice interaction cycles are from the current voice interaction cycle to the past t voice interaction cycles; S12: Selecting the recognition function A1 as the function to be deconstructed, and selecting a voice interaction cycle as the cycle to be deconstructed; Obtaining the sending time of the voice information corresponding to the function to be deconstructed from the period to be deconstructed, and taking the sending time as the standard time, selecting the recognition function corresponding to the voice information whose sending time is closest to the standard time and sent after the standard time from the period to be deconstructed, and using the selected recognition function as the cascade function of the function to be deconstructed based on the period to be deconstructed; At the same time, a recognition function chain list of the period to be deconstructed is constructed according to the recognition functions of all voice information sent by the authorized user in the period to be deconstructed, and all the recognition functions in the recognition function chain list are sorted from left to right in the order of the distance between the sending time of the corresponding voice information and the current time; S13: Select t voice interaction cycles as the cycles to be deconstructed in sequence according to S11 to S12, obtain the cascade functions of the functions to be deconstructed based on the t voice interaction cycles and deduplicate them, mark all the remaining cascade functions after deduplication as B1, B2, ..., Bb, where b≥1, and construct a recognition function linked list of t voice interaction cycles; S14: using the formula Cc=Bc / t to sequentially calculate and obtain the cascade weights C1, C2, ..., Cb of the cascade functions B1, B2, ..., Bb based on the function to be deconstructed, where Bc in the formula refers to each of the cascade functions B1, B2, ..., Bb, and the value of c is 1, 2, ..., b; Generate a cascade function table of the functions to be deconstructed based on the cascade weights of the functions to be deconstructed according to the cascade functions B1, B2, ..., Bb; S15: Obtain the total number D1 of recognition functions to be deconstructed from the recognition functions of all voice messages sent by authorized users in t voice interaction cycles, and calculate the frequency weight E1 of the function to be deconstructed using the formula E1=D1 / D, where D is the total number of all voice messages sent by authorized users in t voice interaction cycles; S16: sequentially selecting recognition functions A2, A3, ..., Aa as functions to be deconstructed, and sequentially generating a cascade function table of recognition functions A2, A3, ..., Aa and their frequency weights according to S13 to S15; S17: Generate a function frequency table according to the frequency weights of the recognition functions A1, A2, ..., Aa; The function deconstruction module transmits the generated function frequency table and the cascade function table of several recognition functions to the recognition matching unit for storage, and transmits the recognition function chain table of t voice interaction cycles to the interaction display unit for storage.
3. A human-computer interaction system based on speech recognition according to claim 1, It is characterized in that In S22, if the similarities between the feature texts of all cascade functions included in the cascade function table and the feature text of the current voice information are less than or equal to P2 after the similarity calculation is completed, the feature texts of all business functions are extracted from top to bottom in sequence according to the arrangement order of all business functions in the function frequency table, and each time the feature text of a business function is extracted, the similarity between the feature text and the feature text of the current voice information is calculated, and after the feature texts of all business functions are completely extracted, the business function corresponding to the feature text participating in the similarity calculation with the largest value is obtained as the function to be confirmed for the current voice information.
4. A human-computer interaction system based on speech recognition according to claim 1, It is characterized in that It also includes a function realization unit, in which the realization logic of all business functions is pre-stored. After receiving the recognition function of the transmitted current voice information, the function realization unit searches for the corresponding realization logic and executes the realization logic.
