Voice interaction method and system based on artificial intelligence

Abstract

The invention discloses a voice interaction method and system based on artificial intelligence, and relates to the field of voice interaction. By analyzing the environment in which the voice interaction system is located, the volume is intelligently adjusted to avoid the uncomfortable experience caused by too high or too low sound; by analyzing the speech speed of new instructions and the similarity between the new instructions and the input corresponding instructions, the instruction analysis index is obtained, and different strategies are adopted for different instruction analysis indexes, so that the instructions can be reasonably handled, thereby improving the intelligence level of the voice interaction system; by analyzing the user's age, region, gender, occupation and user behavior data, the recommended value of each voice package is obtained, so that the appropriate voice package can be intelligently recommended to the user, thereby increasing the user's satisfaction and stickiness.

Description

Technical Field

[0001] The present invention relates to the field of voice interaction, and in particular to a voice interaction method and system based on artificial intelligence. Background Art

[0002] With the popularization of smart devices, such as smartphones, smart speakers, and smart cars, users have a growing demand for convenient interaction methods. Voice interaction, as a natural and efficient interaction method, enables users to easily operate devices when their hands are busy or their vision is limited. However, existing voice interaction systems still have the following defects:

[0003] When the voice interaction system is outputting content, the user may have issued a new command. The existing voice interaction system either ignores the new command or directly executes the new command. The processing method is single and cannot comprehensively analyze the similarity between the user's speech speed and the new command and the command corresponding to the current output content to provide a variety of processing solutions; it cannot intelligently adjust the volume according to the current environment, which may cause the sound to be too loud in places where quietness is required, and too soft in noisy environments; it cannot intelligently recommend suitable voice packages for users;

[0004] To this end, a voice interaction method and system based on artificial intelligence is introduced. Summary of the invention

[0005] In view of this, the present invention provides a voice interaction method and system based on artificial intelligence to solve the problems raised by the above background technology.

[0006] The purpose of the present invention can be achieved by the following technical solution: A voice interaction method based on artificial intelligence comprises the following steps:

[0007] A1: Identify the user's timbre and wake-up command, and wake up the voice interaction system;

[0008] A2: Detect the surrounding environment and intelligently adjust the decibel value of the output voice. The specific process is as follows:

[0009] Monitor the decibel value of the surrounding environment in real time, preset a decibel critical value, if the decibel value of the surrounding environment is less than the decibel critical value, determine that the current environment is a quiet environment, preset a decibel value interval for the decibel value below the decibel critical value, set different decibel value intervals corresponding to different output voice decibel values, match the decibel value interval corresponding to the decibel value of the current environment, and thus obtain the output voice decibel value of the current environment; if the decibel value of the surrounding environment is greater than the decibel critical value, perform the following S1-S3 operations:

[0010] S1: Get the decibel value FB of each sound source in the current environment, the distance JL from the device, and the distance YJ between the user and the device; the device refers to the device on which the voice interaction system is currently deployed, and the distance JL from the device refers to the distance between the sound source and the device on which the voice interaction system is currently deployed, using the formula: Thus, the decibel impact value of each sound source in the current environment is obtained, where a1 and a2 are the decibel value FB of each sound source in the current environment and the weight impact factor corresponding to the distance JL from the device, respectively. The decibel impact value of all people in the current environment is averaged to obtain the average decibel value of the device;

[0011] S2: preset each group of intervals of the distance YJ between the user and the device, set each group of intervals of the distance YJ between the user and the device corresponding to each group of amplification factors, match the interval of the distance YJ between the user and the device corresponding to the distance YJ between the user and the device, so as to obtain the amplification factor corresponding to the distance YJ between the user and the device;

[0012] S3: multiplying the average decibel value of the device by the amplification factor to obtain the comprehensive decibel value of the device, presetting the comprehensive decibel value interval, setting different comprehensive decibel value intervals to correspond to different output voice decibel values, matching the comprehensive decibel value interval corresponding to the comprehensive decibel value, thereby obtaining the output voice decibel value of the current environment;

[0013] A3: Identify the instructions issued by the user and execute them;

[0014] A4: When the voice interaction system is outputting, if it is detected that the user has issued a new instruction, the new instruction is analyzed to obtain the instruction analysis index, and corresponding operations are performed according to the instruction analysis index;

[0015] A5: Calculate the recommendation value of each voice package and recommend the voice package to the user based on the value.

[0016] Furthermore, the specific steps of obtaining the instruction analysis index in A4 are as follows:

[0017] The received new user instructions are decomposed into independent Chinese characters, and the voice duration of each Chinese character and the interval time between two adjacent Chinese characters are obtained by the timestamp marking method; the historical instructions of the user are analyzed to obtain the standard value of the interval time between two adjacent Chinese characters, and the data with the interval time between two adjacent Chinese characters lower than the standard value are extracted to form a fast value set, and each number in the fast value set is subtracted from the standard value of the interval time, and the number in the fast value set is subtracted from the standard value of the interval time, so as to obtain a fast value difference set, and a fast value difference interval is preset, and different fast value difference intervals are set to correspond to different interval influence factors, and the fast value difference intervals where the numbers in the fast value difference set are located are matched, so as to obtain the interval influence factors corresponding to all the numbers in the fast value difference set, and the larger the fast value difference is, the larger the value of the corresponding interval influence factor is, and the number in the fast value difference set is multiplied by the corresponding interval influence factor and then summed, so as to obtain the interval indication value F3;

[0018] Extract the maximum and minimum values ​​of all Chinese character voice durations, subtract the minimum value from the maximum value of the Chinese character voice duration, and obtain the duration difference F1. Use the mean formula to calculate the voice duration of all Chinese characters, and obtain the mean value F2. After normalizing the obtained interval indicator F3, duration difference F1 and mean value F2, enter the formula: Thus, the speech rate index value YS is obtained, where w1, w2 and w3 are weighted influence factors corresponding to the duration difference value F1, the time average value F2 and the interval indication value F3 respectively;

[0019] The received new instructions and the instructions corresponding to the current output content are semantically converted, and the phrases of the converted sentences are extracted. The phrases corresponding to the new instructions are recorded as the new phrase set, and the phrases corresponding to the current output content are recorded as the old phrase set. The word vector model is selected to convert the phrases in the new phrase set and the old phrase set into vectors; the phrases in the new phrase set are used as the row headers of the matrix, and the phrases in the old phrase set are used as the column headers of the matrix. For each phrase in the new phrase set, the cosine similarity is used to calculate the similarity with all the phrases in the old phrase set. Similarity value, fill all the similarity values ​​obtained into the matrix, preset the similarity threshold, classify by row, count the number of similarity values ​​in each row that are higher than the similarity threshold, record it as the row excess value, divide the row excess value of each row by the sum of the row excess values ​​of all rows, so as to obtain the importance ratio of each row, use the mean formula to calculate the similarity value of each row, so as to obtain the row mean of each row, multiply the importance ratio of each row by the row mean of the corresponding row, and add the obtained results, and the final result is used as the similarity index XSD between the new instruction and the instruction corresponding to the current output content;

[0020] The obtained speech speed index value YS and similarity index XSD are normalized and then input into the formula: Thus, the instruction analysis index ZLF is obtained, where z1 and z2 are weight influence factors corresponding to the speech speed index value YS and the similarity index XSD respectively.

[0021] Furthermore, the voice interaction system in A4 processes the obtained instruction analysis index as follows:

[0022] A command analysis index interval is preset. If the command analysis index is greater than the maximum value of the command analysis index interval, the current output is stopped immediately and a new command is executed instead. If the command analysis index is within the command analysis index interval, a question is asked to the user: whether to stop the current output and execute a new user command. If the user answers yes, the new command is executed. If the user answers no, the current output is continued and the new command is executed after the output is completed. If the command analysis index is less than the minimum value of the command analysis index interval, the new command is stored in the cache interval, and after the current output is completed, the user is asked whether to execute the new command. If the user answers yes, the new command is executed. If the user answers no, the new command is moved out of the cache interval and not executed.

[0023] Furthermore, the specific steps of intelligently recommending voice packages to users in A5 are:

[0024] Get the user's age, gender, region and occupation, set the age range, match the user's age range, perform triple screening based on the user's region, gender and age range, and obtain the voice package popularity ranking list 1 after the triple screening from the website background data; at the same time, perform triple screening based on the user's region, gender and occupation, and obtain the voice package popularity ranking list 2 after the triple screening from the website background data;

[0025] Ten voice packets are randomly intercepted from both Ranking 1 and Ranking 2, and the intercepted Ranking 1 and Ranking 2 are deduplicated. The voice packets are numbered and the ranking weight is used to indicate whether a voice packet appears in Ranking 1 or Ranking 2. If a voice packet appears in both Ranking 1 and Ranking 2, the ranking weight is recorded as 2. If a voice packet only appears in Ranking 1 or Ranking 2, the ranking weight is recorded as 1. For voice packets with a ranking weight of 2, Indicates the ranking of the voice package in the ranking list 1, using Indicates the ranking of the voice package in the ranking list 2, u represents the number of the voice package, u=1,2...x, x represents the number of voice packages after deduplication; the weight influence factors corresponding to the preset ranking list 1 and ranking list 2 are used and Multiply the weight factor of ranking list 1 and the weight factor of ranking list 2 respectively and then sum them up, and divide the result by 2 to get the ranking score of the voice package with a ranking number of 2; for the voice package with a ranking number of 1, multiply the ranking of the ranking list where the voice package is located by the weight factor of the corresponding ranking list to get the ranking score of the voice package with a ranking number of 1;

[0026] Obtain the user's behavior data for the deduplicated voice package, including the number of times the deduplicated voice package is used CSZ, the time of the most recent use, and the score PFZ of the voice package; subtract the time of the most recent use of the voice package from the current time of waking up the voice interaction system to obtain the interval usage time JJS of the voice package, and normalize the number of times the voice package is used CSZ, the score PFZ, and the interval usage time JJS and enter them into the formula: Get the usage score of the voice package, where d1, d2 and d3 are the weighted influencing factors corresponding to the number of times the voice package is used CSZ, the score PFZ and the interval usage time JJS respectively;

[0027] The ranking scores and usage scores of the preset voice packages correspond to weight influence factors, the ranking scores and usage scores of the voice packages are normalized, multiplied by the corresponding weight influences and then summed to obtain the recommendation value of the voice package, and the voice package with the highest recommendation value is selected as the user's recommended voice package.

[0028] Furthermore, the artificial intelligence-based voice interaction system includes:

[0029] Wake-up module: built-in microphone matrix, used to collect sound information, including: sound source decibel value, sound source distance, use multiple verification methods to identify and verify the user's voice, and wake up the voice interaction system after successful recognition;

[0030] Volume adjustment module: automatically adjusts the volume of the interactive system according to the sound information of the surrounding environment;

[0031] User interaction module: used for users to issue instructions, and the interactive system recognizes and executes the instructions after receiving them;

[0032] Interrupt management module: used to arrange the execution order of commands when the interactive information is outputting content but a new user command is received at the same time;

[0033] Voice package recommendation module: Calculates the user's recommendation value for each voice package based on the user's region, gender, age range, occupation and historical behavior information, and selects the voice package with the largest recommendation value as the voice package recommended to the user.

[0034] Compared with the prior art, the present invention has the following beneficial effects:

[0035] The present invention sets a decibel critical value, obtains a voice output decibel value directly according to the environmental decibel value for an environment below the decibel critical value, obtains a comprehensive decibel value of the current environment according to the decibel value FB of each sound source, the distance JL from the device, and the distance YJ between the user and the device for an environment above the decibel critical value, determines an amplification factor according to the distance between the user and the device, and finally obtains a voice output allocation value according to the comprehensive allocation value and the amplification factor, thereby enabling a device equipped with a voice interaction system to have a lower volume in a place where quietness is required and a higher volume in a relatively noisy place, thereby realizing intelligent volume adjustment;

[0036] The present invention analyzes the speech speed of a new instruction issued by a user to obtain a speech speed indication value, which can reflect the user's anxiety about the new instruction. By performing similarity analysis on the new instruction and the instruction corresponding to the current output content, a similarity index is obtained, which can judge the similarity between the new instruction and the instruction corresponding to the current output content. If the similarity is too high, the current output content is likely to contain the content that the new instruction needs to output. The instruction analysis index is obtained by combining the similarity index and the speech speed indication value, so as to reasonably handle the new instruction and improve the intelligence of the voice interaction system.

[0037] The present invention obtains two ranking lists by screening users according to age, region, gender and occupation, obtains ranking scores of voice packages by setting weight factors of the two ranking lists and analyzing the rankings of the two ranking lists, and obtains usage scores of voice packages by analyzing user behavior data at the same time. Finally, the usage scores and ranking scores of voice packages are comprehensively considered to obtain the recommendation value of each voice package, so that suitable voice packages can be intelligently recommended to users, thereby increasing user satisfaction and stickiness. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Further details, features and advantages of the present application are disclosed in the following description of exemplary embodiments in conjunction with the accompanying drawings, in which:

[0039] Figure 1 It is a principle block diagram of the present invention. DETAILED DESCRIPTION

[0040] Several embodiments of the present application will be described in more detail below with reference to the accompanying drawings so that those skilled in the art can implement the present application. The present application can be embodied in many different forms and purposes and should not be limited to the embodiments described herein. These embodiments are provided to make the present application comprehensive and complete, and to fully convey the scope of the present application to those skilled in the art. The embodiments do not limit the present application.

[0041] Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by those of ordinary skill in the art to which this application belongs. It will be further understood that terms such as those defined in commonly used dictionaries should be interpreted as having a meaning consistent with their meaning in the relevant art and / or the context of this specification, and will not be interpreted in an idealized or overly formal sense unless explicitly defined as such herein.

[0042] See also Figure 1 As shown in the figure, the voice interaction method based on artificial intelligence includes the following five steps:

[0043] A1: Identify the user's tone and wake-up command, and wake up the voice interaction system.

[0044] A2: Detect the surrounding environment and adjust the decibel value of the output voice;

[0045] Monitor the decibel value of the surrounding environment in real time, preset a decibel critical value, if the decibel value of the surrounding environment is less than the decibel critical value, determine that the current environment is a quiet environment, preset a decibel value interval for the decibel value below the decibel critical value, set different decibel value intervals corresponding to different output voice decibel values, match the decibel value interval corresponding to the decibel value of the current environment, and thus obtain the output voice decibel value of the current environment; if the decibel value of the surrounding environment is greater than the decibel critical value, perform the following S1-S3 operations:

[0046] By presetting the decibel value range for quiet environments below the decibel threshold and corresponding different output voice decibel values, the system can accurately match the appropriate output voice decibel according to the quietness of the actual environment, avoiding excessive voice volume that destroys the quiet atmosphere, ensuring that the sound output during voice interaction can be clearly heard by users and can be coordinated with the quiet environment, thus improving the user's interactive experience in quiet scenes;

[0047] S1: Get the decibel value FB of each sound source in the current environment, the distance JL from the device, and the distance YJ between the user and the device; the device refers to the device on which the voice interaction system is currently deployed, and the distance JL from the device refers to the distance between the sound source and the device on which the voice interaction system is currently deployed, using the formula: Thus, the decibel impact value of each sound source in the current environment is obtained, where a1 and a2 are the decibel value FB of each sound source in the current environment and the weight impact factor corresponding to the distance JL from the device, respectively. The decibel impact value of all people in the current environment is averaged to obtain the average decibel value of the device;

[0048] S2: preset each group of intervals of the distance YJ between the user and the device, set each group of intervals of the distance YJ between the user and the device corresponding to each group of amplification factors, match the interval of the distance YJ between the user and the device corresponding to the distance YJ between the user and the device, so as to obtain the amplification factor corresponding to the distance YJ between the user and the device;

[0049] The distance between different individuals and the device is fully considered, and corresponding amplification factors are set for different distance intervals. The final output voice decibel value is determined in combination with the device average decibel value, thus achieving personalized volume adaptation.

[0050] S3: Multiply the average decibel value of the device by the amplification factor to obtain the comprehensive decibel value of the device, preset the comprehensive decibel value range, set different comprehensive decibel value ranges to correspond to different output voice decibel values, match the comprehensive decibel value corresponding to the comprehensive decibel value range, and thus obtain the output voice decibel value of the current environment.

[0051] A3: Used to identify and execute instructions issued by users.

[0052] A4: When the voice interaction system is outputting, if it is detected that the user has issued a new instruction, the new instruction is analyzed to obtain the instruction analysis index, and corresponding operations are performed according to the instruction analysis index;

[0053] The received new user instructions are decomposed into independent Chinese characters, and the voice duration of each Chinese character and the interval time between two adjacent Chinese characters are obtained by the timestamp marking method; the historical instructions of the user are analyzed to obtain the standard value of the interval time between two adjacent Chinese characters, and the data with the interval time between two adjacent Chinese characters lower than the standard value are extracted to form a fast value set, and each number in the fast value set is subtracted from the standard value of the interval time, and the number in the fast value set is subtracted from the standard value of the interval time, so as to obtain a fast value difference set, and a fast value difference interval is preset, and different fast value difference intervals are set to correspond to different interval influence factors, and the fast value difference intervals where the numbers in the fast value difference set are located are matched, so as to obtain the interval influence factors corresponding to all the numbers in the fast value difference set, and the larger the fast value difference is, the larger the value of the corresponding interval influence factor is, and the number in the fast value difference set is multiplied by the corresponding interval influence factor and then summed, so as to obtain the interval indication value F3;

[0054] By analyzing the user's historical commands to determine the standard value of the interval time, the language expression habits of different users are fully taken into account. Each user has uniqueness in terms of speaking rhythm, word spacing, etc., and using this standard value as a reference to analyze the speech speed-related indicators of new commands can better fit the normal expression style of individual users, avoid ignoring individual differences due to unified standards, and make the system's judgment of the user's speech speed more in line with the actual situation, thereby improving the accuracy of understanding user commands;

[0055] Extract the maximum and minimum values ​​of all Chinese character voice durations, subtract the minimum value from the maximum value of the Chinese character voice duration, and obtain the duration difference F1. Use the mean formula to calculate the voice duration of all Chinese characters, and obtain the mean value F2. After normalizing the obtained interval indicator F3, duration difference F1 and mean value F2, enter the formula: Thus, the speech rate index value YS is obtained, where w1, w2 and w3 are weighted influence factors corresponding to the duration difference value F1, the time average value F2 and the interval indication value F3 respectively;

[0056] The duration difference F1, the average value F2 and the interval indicator F3 are integrated, and the speech rate index value YS is calculated by assigning different weights to the influencing factors, thus achieving a comprehensive and integrated evaluation of the speech rate from multiple dimensions.

[0057] The received new instructions and the instructions corresponding to the current output content are semantically converted, and the phrases of the converted sentences are extracted. The phrases corresponding to the new instructions are recorded as the new phrase set, and the phrases corresponding to the current output content are recorded as the old phrase set. The word vector model is selected to convert the phrases in the new phrase set and the old phrase set into vectors; the phrases in the new phrase set are used as the row headers of the matrix, and the phrases in the old phrase set are used as the column headers of the matrix. For each phrase in the new phrase set, the cosine similarity is used to calculate the similarity with all the phrases in the old phrase set. Similarity value, fill all the similarity values ​​obtained into the matrix, preset the similarity threshold, classify by row, count the number of similarity values ​​in each row that are higher than the similarity threshold, record it as the row excess value, divide the row excess value of each row by the sum of the row excess values ​​of all rows, so as to obtain the importance ratio of each row, use the mean formula to calculate the similarity value of each row, so as to obtain the row mean of each row, multiply the importance ratio of each row by the row mean of the corresponding row, and add the obtained results, and the final result is used as the similarity index XSD between the new instruction and the instruction corresponding to the current output content;

[0058] By performing semantic conversion on the new instruction and the instruction corresponding to the current output content, extracting phrases and converting them into vectors, and then using cosine similarity to build a similarity matrix, the semantic correlation between the two instructions can be deeply and accurately analyzed; if the new instruction is highly correlated with the instruction corresponding to the current output content, the current output content is likely to contain the content of the new instruction, and the execution order of the new instruction can be postponed;

[0059] The obtained speech speed index value YS and similarity index XSD are normalized and then input into the formula: Thus, the instruction analysis index ZLF is obtained, where z1 and z2 are weight influence factors corresponding to the speech speed index value YS and the similarity index XSD respectively.

[0060] A command analysis index interval is preset. If the command analysis index is greater than the maximum value of the command analysis index interval, the current output is stopped immediately and a new command is executed instead. If the command analysis index is within the command analysis index interval, a question is asked to the user: whether to stop the current output and execute a new user command. If the user answers yes, the new command is executed. If the user answers no, the current output is continued and the new command is executed after the output is completed. If the command analysis index is less than the minimum value of the command analysis index interval, the new command is stored in the cache interval, and after the current output is completed, the user is asked whether to execute the new command. If the user answers yes, the new command is executed. If the user answers no, the new command is removed from the cache interval and not executed.

[0061] The execution order of instructions is determined based on the comparison result between the instruction analysis index and the preset interval, avoiding confusion and disorder in instruction processing. The system will not blindly interrupt or overlap instruction execution, but will arrange it in order according to reasonable rules, effectively reducing the time waste and resource consumption caused by instruction conflicts, and improving the overall efficiency of task execution, especially in multi-tasking concurrency or complex instruction interaction scenarios.

[0062] A5: Calculate the recommendation value of each voice package and recommend the voice package to the user based on the value;

[0063] Get the user's age, gender, region and occupation, set the age range, match the user's age range, perform triple screening based on the user's region, gender and age range, and obtain the voice package popularity ranking list 1 after the triple screening from the website background data; at the same time, perform triple screening based on the user's region, gender and occupation, and obtain the voice package popularity ranking list 2 after the triple screening from the website background data;

[0064] Comprehensively obtain multi-dimensional information such as the user's age, gender, region and occupation, and perform multiple screening from different angles to generate a voice pack popularity ranking list; this fully considers the impact of different factors on the user's voice pack selection preferences. For example, users of different age ranges, different regional cultural backgrounds and different occupational characteristics often have different preferences for voice packs. This method can more accurately discover voice packs that meet the user's potential preferences and avoid the one-sidedness of single-factor recommendations;

[0065] Ten voice packets are randomly intercepted from both Ranking 1 and Ranking 2, and the intercepted Ranking 1 and Ranking 2 are deduplicated. The voice packets are numbered and the ranking weight is used to indicate whether a voice packet appears in Ranking 1 or Ranking 2. If a voice packet appears in both Ranking 1 and Ranking 2, the ranking weight is recorded as 2. If a voice packet only appears in Ranking 1 or Ranking 2, the ranking weight is recorded as 1. For voice packets with a ranking weight of 2, Indicates the ranking of the voice package in the ranking list 1, using Indicates the ranking of the voice package in the ranking list 2, u represents the number of the voice package, u=1,2...x, x represents the number of voice packages after deduplication; the weight influence factors corresponding to the preset ranking list 1 and ranking list 2 are used and Multiply the weight factor of ranking list 1 and the weight factor of ranking list 2 respectively and then sum them up, and divide the result by 2 to get the ranking score of the voice package with a ranking number of 2; for the voice package with a ranking number of 1, multiply the ranking of the ranking list where the voice package is located by the weight factor of the corresponding ranking list to get the ranking score of the voice package with a ranking number of 1;

[0066] For voice packs that appear differently in different rankings, we use a detailed ranking score calculation method to accurately measure their advantages in different dimensions. Whether it is a voice pack that is popular in multiple rankings or a voice pack that only performs well in one aspect, it can be reasonably evaluated to ensure that the recommended voice pack is the most advantageous and most suitable choice for the user after comprehensive consideration;

[0067] Obtain the user's behavior data for the deduplicated voice package, including the number of times the deduplicated voice package is used CSZ, the time of the most recent use, and the score PFZ of the voice package; subtract the time of the most recent use of the voice package from the current time of waking up the voice interaction system to obtain the interval usage time JJS of the voice package, and normalize the number of times the voice package is used CSZ, the score PFZ, and the interval usage time JJS and enter them into the formula: Get the usage score of the voice package, where d1, d2 and d3 are the weighted influencing factors corresponding to the number of times the voice package is used CSZ, the score PFZ and the interval usage time JJS respectively;

[0068] The ranking scores and usage scores of the preset voice packages correspond to weight influence factors, the ranking scores and usage scores of the voice packages are normalized, multiplied by the corresponding weight influences and then summed to obtain the recommendation value of the voice package, and the voice package with the highest recommendation value is selected as the user's recommended voice package.

[0069] The voice interaction system based on artificial intelligence includes: wake-up module, volume adjustment module, user interaction module, interruption management module and voice package recommendation module;

[0070] Wake-up module: built-in microphone matrix, used to collect sound information, including: sound source decibel value, sound source distance, use multiple verification methods to identify and verify the user's voice, and wake up the voice interaction system after successful recognition;

[0071] Volume adjustment module: automatically adjusts the volume of the interactive system according to the sound information of the surrounding environment;

[0072] User interaction module: used for users to issue instructions, and the interactive system recognizes and executes the instructions after receiving them;

[0073] Interrupt management module: used to arrange the execution order of commands when the interactive information is outputting content but a new user command is received at the same time;

[0074] Voice package recommendation module: Calculates the user's recommendation value for each voice package based on the user's region, gender, age range, occupation and historical behavior information, and selects the voice package with the largest recommendation value as the voice package recommended to the user.

[0075] The preferred embodiments of the present invention disclosed above are only used to help explain the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to only specific implementation methods. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and use the present invention well. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. A voice interaction method based on artificial intelligence, characterized in that: include: A1: Identify the user's timbre and wake-up command, and wake up the voice interaction system; A2: Detect the surrounding environment and adjust the decibel value of the output voice. The specific process is as follows: Monitor the decibel value of the surrounding environment in real time, preset a decibel critical value, if the decibel value of the surrounding environment is less than the decibel critical value, determine that the current environment is a quiet environment, preset a decibel value interval for the decibel value below the decibel critical value, set different decibel value intervals corresponding to different output voice decibel values, match the decibel value interval corresponding to the decibel value of the current environment, and thus obtain the output voice decibel value of the current environment; if the decibel value of the surrounding environment is greater than the decibel critical value, perform the following S1-S3 operations: S1: Get the decibel value FB of each sound source in the current environment, the distance JL from the device, and the distance YJ between the user and the device; the device refers to the device on which the voice interaction system is currently deployed, and the distance JL from the device refers to the distance between the sound source and the device on which the voice interaction system is currently deployed, using the formula: Thus, the decibel impact value of each sound source in the current environment is obtained, where a1 and a2 are the decibel value FB of each sound source in the current environment and the weight impact factor corresponding to the distance JL from the device, respectively. The decibel impact value of all people in the current environment is averaged to obtain the average decibel value of the device; S2: preset each group of intervals of the distance YJ between the user and the device, set each group of intervals of the distance YJ between the user and the device corresponding to each group of amplification factors, match the interval of the distance YJ between the user and the device corresponding to the distance YJ between the user and the device, so as to obtain the amplification factor corresponding to the distance YJ between the user and the device; S3: multiplying the average decibel value of the device by the amplification factor to obtain the comprehensive decibel value of the device, presetting the comprehensive decibel value interval, setting different comprehensive decibel value intervals to correspond to different output voice decibel values, matching the comprehensive decibel value interval corresponding to the comprehensive decibel value, thereby obtaining the output voice decibel value of the current environment; A3: Identify the instructions issued by the user and execute them; A4: When the voice interaction system is outputting, if it is detected that the user has issued a new instruction, the new instruction is analyzed to obtain the instruction analysis index, and corresponding operations are performed according to the instruction analysis index; The specific steps of obtaining the instruction analysis index in A4 are as follows: Decompose the received new user instructions into independent Chinese characters, obtain the voice duration of each Chinese character and the interval time between two adjacent Chinese characters; analyze the user's historical instructions to obtain the standard value of the interval time between two adjacent Chinese characters, extract the data with the interval time between two adjacent Chinese characters lower than the standard value to form a fast value set, subtract each number in the fast value set from the standard value of the interval time, subtract the number in the fast value set from the standard value of the interval time, thereby obtaining a fast value difference set, preset a fast value difference interval, set different interval influence factors corresponding to different fast value difference intervals, match the fast value difference intervals where the numbers in the fast value difference set are located, thereby obtaining the interval influence factors corresponding to all the numbers in the fast value difference set, multiply the numbers in the fast value difference set by the corresponding interval influence factors and add them together, thereby obtaining the interval indication value F3; Extract the maximum and minimum values ​​of all Chinese character voice durations, subtract the minimum value from the maximum value of the Chinese character voice duration, and obtain the duration difference F1. Use the mean formula to calculate the voice duration of all Chinese characters, and obtain the mean value F2. After normalizing the obtained interval indicator F3, duration difference F1 and mean value F2, enter the formula: Thus, the speech rate index value YS is obtained, where w1, w2 and w3 are weighted influence factors corresponding to the duration difference value F1, the time average value F2 and the interval indication value F3 respectively; Perform semantic conversion on the received new instructions and the instructions corresponding to the current output content, extract the phrases of the converted sentences, record the phrases corresponding to the new instructions as the new phrase set, record the phrases corresponding to the current output content as the old phrase set, and convert the phrases in the new phrase set and the old phrase set into vectors; use the phrases in the new phrase set as the row headers of the matrix, and use the phrases in the old phrase set as the column headers of the matrix; for each phrase in the new phrase set, calculate the similarity value with all the phrases in the old phrase set, fill all the obtained similarity values ​​into the matrix, preset a similarity threshold, classify by row, count the number of similarity values ​​in each row that are higher than the similarity threshold, record it as the row excess value, divide the row excess value of each row by the sum of the row excess values ​​of all rows, so as to obtain the importance ratio of each row, use the mean formula to calculate the similarity value of each row, so as to obtain the row mean of each row, multiply the importance ratio of each row by the row mean of the corresponding row, add the obtained results, and the final result is used as the similarity index XSD of the new instruction and the instruction corresponding to the current output content; The obtained speech speed index value YS and similarity index XSD are normalized and then input into the formula: Thus, the instruction analysis index ZLF is obtained, where z1 and z2 are the weighted influence factors corresponding to the speech speed index value YS and the similarity index XSD respectively; A5: Calculate the recommendation value of each voice package and recommend the voice package to the user based on the value.

2. The artificial intelligence-based voice interaction method according to claim 1, characterized in that: The voice interaction system in A4 processes the obtained command analysis index as follows: A command analysis index interval is preset. If the command analysis index is greater than the maximum value of the command analysis index interval, the current output is stopped immediately and a new command is executed instead. If the command analysis index is within the command analysis index interval, a question is asked to the user: whether to stop the current output and execute a new user command. If the user answers yes, the new command is executed. If the user answers no, the current output is continued and the new command is executed after the output is completed. If the command analysis index is less than the minimum value of the command analysis index interval, the new command is stored in the cache interval, and after the current output is completed, the user is asked whether to execute the new command. If the user answers yes, the new command is executed. If the user answers no, the new command is moved out of the cache interval and not executed.

3. The artificial intelligence-based voice interaction method according to claim 1, characterized in that: The specific steps of intelligently recommending voice packages to users in A5 are: Get the user's age, gender, region and occupation, set the age range, match the user's age range, perform triple screening based on the user's region, gender and age range, and obtain the voice package popularity ranking list 1 after the triple screening from the website background data; at the same time, perform triple screening based on the user's region, gender and occupation, and obtain the voice package popularity ranking list 2 after the triple screening from the website background data; Ten voice packets are randomly intercepted from both Ranking 1 and Ranking 2, and the intercepted Ranking 1 and Ranking 2 are deduplicated. The voice packets are numbered and the ranking weight is used to indicate whether a voice packet appears in Ranking 1 or Ranking 2. If a voice packet appears in both Ranking 1 and Ranking 2, the ranking weight is recorded as 2. If a voice packet only appears in Ranking 1 or Ranking 2, the ranking weight is recorded as 1. For voice packets with a ranking weight of 2, Indicates the ranking of the voice package in the ranking list 1, using Indicates the ranking of the voice package in the ranking list 2, u represents the number of the voice package, u=1,2...x, x represents the number of voice packages after deduplication; the weight influence factors corresponding to the preset ranking list 1 and ranking list 2 are used and Multiply the weight factor of ranking list 1 and the weight factor of ranking list 2 respectively, then sum the results and divide the result by 2 to get the ranking score of the voice package with a ranking weight of 2; For a voice package with a ranking weight of 1, the ranking of the voice package in the ranking list is multiplied by the weight factor of the corresponding ranking list to obtain the ranking score of the voice package with a ranking weight of 1; Obtain the user's behavior data for the deduplicated voice package, including the number of times the deduplicated voice package is used CSZ, the time of the most recent use, and the score PFZ of the voice package; subtract the time of the most recent use of the voice package from the current time of waking up the voice interaction system to obtain the interval usage time JJS of the voice package, and normalize the number of times the voice package is used CSZ, the score PFZ, and the interval usage time JJS and enter them into the formula: Get the usage score of the voice package, where d1, d2 and d3 are the weighted influencing factors corresponding to the number of times the voice package is used CSZ, the score PFZ and the interval usage time JJS respectively; The ranking scores and usage scores of the preset voice packages correspond to weight influence factors, the ranking scores and usage scores of the voice packages are normalized, multiplied by the corresponding weight influences and then summed to obtain the recommendation value of the voice package, and the voice package with the highest recommendation value is selected as the user's recommended voice package.

4. A control system based on the artificial intelligence-based voice interaction method according to any one of claims 1 to 3, characterized in that: Contains the following modules: Wake-up module: built-in microphone matrix, used to collect sound information, including: sound source decibel value, sound source distance, use multiple verification methods to identify and verify the user's voice, and wake up the voice interaction system after successful recognition; Volume adjustment module: automatically adjusts the volume of the interactive system according to the sound information of the surrounding environment; User interaction module: used for users to issue instructions, and the interactive system recognizes and executes the instructions after receiving them; Interrupt management module: used to arrange the execution order of commands when the interactive information is outputting content but a new user command is received at the same time; Voice package recommendation module: Calculates the user's recommendation value for each voice package based on the user's region, gender, age range, occupation and historical behavior information, and selects the voice package with the largest recommendation value as the voice package recommended to the user.

Images