Voice control method and system applied to multiple servers and storage medium

Abstract

The invention relates to a voice control method and system applied to multiple servers and a storage medium, and the method comprises the steps: U1, enabling a user to carry out voice input, obtaining the data information of voice sequences of the users of multiple groups of servers in real time based on a voice receiving sensor of the multiple servers, carrying out the preprocessing and feature extraction, and obtaining a voice sequence of the user; obtaining data information of the feature vectors of the voice sequences of the users of the multiple groups of servers; and U2, based on the data information of the feature vectors of the voice sequences of the users of the multiple groups of servers, optimizing the feature vectors of the voice sequences of the users by adopting an improved whale optimization algorithm to obtain the optimized data information of the feature vectors of the voice sequences of the users of the multiple groups of servers. According to the invention, the input voice can be accurately processed, the quality of the output voice is improved, the voice received by multiple servers can be processed at the same time, and the operation efficiency of the system is improved.

Description

Technical Field

[0001] The present invention relates to the field of language control technology, and in particular to a voice control method, system and storage medium applied to multiple servers. Background Art

[0002] With the continuous development of science and technology, voice recognition has become an indispensable part of people's daily life. Technologies such as voice chat, voice control of home appliances and voice recognition door opening are constantly bringing convenience to people's lives. However, in the process of voice input, it will carry a lot of noise or white noise, resulting in a large error between the output voice and the desired input voice, which is not matched with the precise system. After the match is successful, the desired effect cannot be obtained. Therefore, how to accurately control the output of voice and improve the output quality of voice has become a problem that we need to solve urgently.

[0003] In the prior art, a Chinese patent (application number: 202110228966.0, publication number: CN 113035194B) discloses a voice control method, a display device and a server, which sends the voice text to the server so that the server performs error correction processing on the voice text to obtain the corrected voice text; splits the corrected voice text into several participles, and annotates the participles with attributes; matches the attribute-annotated participles with preset templates to determine alternative dynamic interface word texts and static interface word texts; analyzes the alternative dynamic interface word texts and static interface word texts to determine standardized dynamic interface word texts and standardized static interface word texts; based on the word segmentation results and attribute annotation results, searches for the corrected voice in the preset business classification template A business classification template matching the text; when a matching business classification template is found, determining the first control business corresponding to the business classification template matching the corrected speech text; when a matching business classification template is not found, determining the second control business from the candidate business corresponding to the corrected speech text based on the attribute annotation result and the word segmentation result; performing semantic analysis on the corrected speech text according to the first control business or the second control business to obtain a semantic analysis result; sending the standardized interface word text and the semantic analysis result to the controller, the standardized interface word text includes a standardized dynamic interface word text and a standardized static interface word text. This solution does not accurately process the input speech. Summary of the invention

[0004] In view of the above deficiencies in the prior art, the present invention provides a voice control method, system and storage medium applied to multiple servers, which can not only accurately process the input voice and improve the quality of the output voice, but also process the voices received by multiple servers at the same time, thereby improving the operating efficiency of the system.

[0005] In order to achieve the above-mentioned purpose and other related purposes, the technical solution provided by the present invention is as follows:

[0006] A voice control method applied to multiple servers, the method comprising:

[0007] U1. The user performs voice input, and the voice receiving sensors based on multiple servers obtain the data information of the voice sequences of users of multiple groups of servers in real time, and perform preprocessing and feature extraction to obtain the data information of the feature vectors of the voice sequences of users of multiple groups of servers;

[0008] U2. Based on the data information of the feature vectors of the user's voice sequence of the multiple groups of servers, the feature vectors of the user's voice sequence are optimized using the improved White Whale optimization algorithm to obtain the data information of the feature vectors of the user's voice sequence of the optimized multiple groups of servers;

[0009] U3. Based on the data information of the feature vectors of the user's voice sequence of the optimized multiple groups of servers, a user's language model is constructed, and the feature vectors of the user's voice sequence are analyzed to obtain data information of the feature values ​​of the user's language;

[0010] U4. Based on the data information of the characteristic value of the user's language, construct the quality index function H of the user's language, characterize the quality of the user's language, obtain the quality characterization value of the user's language, set a preset threshold, if the quality characterization value of the user's language is less than the preset threshold, the voice requirements are met, and output the data information of the user's voice. If it is greater than the preset threshold, the voice requirements are not met, and return to step U2.

[0011] Further, in step U4, the quality index function H of the user's language is,

[0012] Among them, x i is the data information of the i-th feature value of the user's language, p is the complexity function of the language, λ i is the weight coefficient of the i-th eigenvalue of the user language, and n is the sample size.

[0013] Furthermore, the complexity function p of the language is,

[0014]

[0015] Among them, x i It is the data information of the i-th feature value of the user's language.

[0016] Furthermore, the weight coefficient λ of the i-th feature value of the user language i The constraints are:

[0017]

[0018] Where n is the sample size.

[0019] Furthermore, in step U2, the optimization of the feature vector of the user's speech sequence by using the improved White Whale optimization algorithm includes:

[0020] U21. Based on the data information of the feature vectors of the speech sequences of the users of the multiple groups of servers, the beluga population is initialized, the population parameters and the maximum number of iterations T are determined, and the data information of the initialized beluga population is obtained;

[0021] U22. Based on the data information of the initialized beluga population, establish the fitness function Q of the population individuals,

[0022]

[0023] Among them, y is the data information of the initialized beluga population, α 1 , α 2 and α 3 is the balance factor of the beluga population, T is the maximum number of iterations, B 0 is a random number in the interval (0,1), and the fitness value of the individual in the population is calculated to obtain the data information of the fitness value of the individual in the population;

[0024] U23. Based on the data information of the fitness values ​​of the individuals in the population, establish a target optimization function W,

[0025]

[0026] Among them, z is the data information of the fitness value of the population individual, β 1 , β 2 and β 3 A random constant parameter in the interval (0,1) is used to optimize the feature vector of the user's speech sequence, and data information of the feature vectors of the speech sequences of the users of the optimized multiple groups of servers is obtained.

[0027] Furthermore, the balance factor α of the beluga population individuals 1 , α 2 and α 3 The constraints are:

[0028]

[0029] Furthermore, in step U3, the construction of the user's language model and the analysis of the feature vector of the user's speech sequence include:

[0030] U31. Based on the data information of the feature vectors of the voice sequences of the users of the optimized multiple groups of servers, a data set of feature vectors of the voice sequences of the users of the optimized multiple groups of servers is constructed;

[0031] U32. Based on the data set of feature vectors of the user's voice sequence of the optimized multiple groups of servers, establish the feature value function R of the user's language,

[0032]

[0033] Among them, r is the dataset of feature vectors of speech sequences of users of multiple groups of servers after optimization, δ 1 , δ 2 and δ 3 is the weight coefficient of the feature vector of the speech sequence;

[0034] U33. Based on the characteristic value function R of the user's language, the characteristic vector of the user's speech sequence is analyzed to obtain data information of the characteristic value of the user's language.

[0035] Furthermore, the weight coefficient δ of the feature vector of the speech sequence 1 , δ 2 and δ 3 for,

[0036]

[0037]

[0038] Wherein, r is a dataset of feature vectors of speech sequences of users of multiple groups of servers after optimization.

[0039] In order to achieve the above-mentioned object and other related objects, the present invention also provides a system for implementing any of the above-mentioned voice control methods applied to multiple servers, the system comprising:

[0040] The data acquisition module is used to obtain data information of speech sequences of users of multiple groups of servers in real time based on speech receiving sensors of multiple servers, and to perform preprocessing and feature extraction to obtain data information of feature vectors of speech sequences of users of multiple groups of servers;

[0041] A voice data optimization module, connected to the data acquisition module, is used to optimize the feature vector of the user's voice sequence using an improved White Whale optimization algorithm to obtain data information of the feature vector of the voice sequence of users of multiple groups of servers after optimization;

[0042] A language data feature value obtaining module, connected to the speech data optimization module, is used to construct a user's language model, analyze the feature vector of the user's speech sequence, and obtain data information of the feature value of the user's language;

[0043] The speech control output module is connected to the language data feature value obtaining module, and is used to construct a quality index function H of the user's language, characterize the quality of the user's language, obtain the quality characterization value of the user's language, set a preset threshold, and if the quality characterization value of the user's language is less than the preset threshold, the speech requirements are met, and the data information of the user's speech is output; if it is greater than the preset threshold, the speech requirements are not met and re-optimization is performed.

[0044] In order to achieve the above-mentioned object and other related objects, the present invention also provides a computer-readable storage medium, on which is stored a computer program programmed or configured to execute any one of the voice control methods applied to multiple servers.

[0045] The present invention has the following positive effects:

[0046] 1. The present invention optimizes the feature vector of the user's voice sequence by adopting an improved White Whale optimization algorithm, and analyzes the feature vector of the user's voice sequence by combining the construction of the user's language model. It can not only accurately process the input voice and improve the quality of the output voice, but also process the voices received by multiple servers at the same time, thereby improving the operating efficiency of the system.

[0047] 2. The present invention characterizes the quality of the user's language by constructing a quality index function H of the user's language and sets a preset threshold. If the quality characterization value of the user's language is less than the preset threshold, the speech requirement is met and the data information of the user's speech is output; if it is greater than the preset threshold, the speech requirement is not met and re-optimization is performed. This not only enables precise control of the input and output of the speech of the multi-server system to ensure the quality of the speech output, but also the whole process is highly robust and can improve the accuracy of the training method through continuous speech input. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 It is a schematic diagram of the method flow of the present invention;

[0049] Figure 2 It is a schematic diagram of the flow chart of the improved Beluga optimization algorithm of the present invention;

[0050] Figure 3 A schematic diagram of the process of building a user's language model according to the present invention;

[0051] Figure 4 It is a schematic diagram of the system framework of the present invention. DETAILED DESCRIPTION

[0052] The following is a description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, the description of well-known functions and structures is omitted in the following description.

[0053] Example 1: Figure 1 As shown, a voice control method applied to multiple servers, the method comprising:

[0054] U1. The user performs voice input, and the voice receiving sensors based on multiple servers obtain the data information of the voice sequences of users of multiple groups of servers in real time, and perform preprocessing and feature extraction to obtain the data information of the feature vectors of the voice sequences of users of multiple groups of servers;

[0055] U2. Based on the data information of the feature vectors of the user's voice sequence of the multiple groups of servers, the feature vectors of the user's voice sequence are optimized using the improved White Whale optimization algorithm to obtain the data information of the feature vectors of the user's voice sequence of the optimized multiple groups of servers;

[0056] U3. Based on the data information of the feature vectors of the user's voice sequence of the optimized multiple groups of servers, a user's language model is constructed, and the feature vectors of the user's voice sequence are analyzed to obtain data information of the feature values ​​of the user's language;

[0057] U4. Based on the data information of the characteristic value of the user's language, construct the quality index function H of the user's language, characterize the quality of the user's language, obtain the quality characterization value of the user's language, set a preset threshold, if the quality characterization value of the user's language is less than the preset threshold, the voice requirements are met, and output the data information of the user's voice. If it is greater than the preset threshold, the voice requirements are not met, and return to step U2.

[0058] In this embodiment, if Figure 2 As shown, in step U2, the use of the improved White Whale optimization algorithm to optimize the feature vector of the user's speech sequence includes:

[0059] U21. Based on the data information of the feature vectors of the speech sequences of the users of the multiple groups of servers, the beluga population is initialized, the population parameters and the maximum number of iterations T are determined, and the data information of the initialized beluga population is obtained;

[0060] U22. Based on the data information of the initialized beluga population, establish the fitness function Q of the population individuals,

[0061]

[0062] Among them, y is the data information of the initialized beluga population, α 1 , α 2 and α 3 is the balance factor of the beluga population, T is the maximum number of iterations, B 0 is a random number in the interval (0,1), and the fitness value of the individual in the population is calculated to obtain the data information of the fitness value of the individual in the population;

[0063] U23. Based on the data information of the fitness values ​​of the individuals in the population, establish a target optimization function W,

[0064]

[0065] Among them, z is the data information of the fitness value of the population individual, β 1 , β 2 and β 3 A random constant parameter in the interval (0,1) is used to optimize the feature vector of the user's speech sequence, and data information of the feature vectors of the speech sequences of the users of the optimized multiple groups of servers is obtained.

[0066] In this embodiment, the balance factor α of the beluga population individuals is 1 , α 2 and α 3 The constraints are:

[0067]

[0068] In this embodiment, if Figure 3 As shown, in step U3, the construction of the user's language model and the analysis of the feature vector of the user's speech sequence include:

[0069] U31. Based on the data information of the feature vectors of the voice sequences of the users of the optimized multiple groups of servers, a data set of feature vectors of the voice sequences of the users of the optimized multiple groups of servers is constructed;

[0070] U32. Based on the data set of feature vectors of the user's voice sequence of the optimized multiple groups of servers, establish the feature value function R of the user's language,

[0071]

[0072] Among them, r is the dataset of feature vectors of speech sequences of users of multiple groups of servers after optimization, δ 1 , δ 2 and δ 3is the weight coefficient of the feature vector of the speech sequence;

[0073] U33. Based on the characteristic value function R of the user's language, the characteristic vector of the user's speech sequence is analyzed to obtain data information of the characteristic value of the user's language.

[0074] In this embodiment, the weight coefficient δ of the feature vector of the speech sequence is 1 , δ 2 and δ 3 for,

[0075]

[0076]

[0077] Wherein, r is a dataset of feature vectors of speech sequences of users of multiple groups of servers after optimization.

[0078] In this embodiment, an improved Beluga optimization algorithm is used to optimize the feature vector of the user's voice sequence, and combined with building the user's language model, the feature vector of the user's voice sequence is analyzed. This can not only accurately process the input voice and improve the quality of the output voice, but also process the voice received by multiple servers at the same time, thereby improving the operating efficiency of the system.

[0079] Embodiment 2: Based on the voice control method applied to multiple servers in Embodiment 1, the present invention is further illustrated and described below.

[0080] like Figure 1 As shown, a voice control method applied to multiple servers, the method comprising:

[0081] U1. The user performs voice input, and the voice receiving sensors based on multiple servers obtain the data information of the voice sequences of users of multiple groups of servers in real time, and perform preprocessing and feature extraction to obtain the data information of the feature vectors of the voice sequences of users of multiple groups of servers;

[0082] U2. Based on the data information of the feature vectors of the user's voice sequence of the multiple groups of servers, the feature vectors of the user's voice sequence are optimized using the improved White Whale optimization algorithm to obtain the data information of the feature vectors of the user's voice sequence of the optimized multiple groups of servers;

[0083] U3. Based on the data information of the feature vectors of the user's voice sequence of the optimized multiple groups of servers, a user's language model is constructed, and the feature vectors of the user's voice sequence are analyzed to obtain data information of the feature values ​​of the user's language;

[0084] U4. Based on the data information of the characteristic value of the user's language, construct the quality index function H of the user's language, characterize the quality of the user's language, obtain the quality characterization value of the user's language, set a preset threshold, if the quality characterization value of the user's language is less than the preset threshold, the voice requirements are met, and output the data information of the user's voice. If it is greater than the preset threshold, the voice requirements are not met, and return to step U2.

[0085] In this embodiment, in step U4, the quality index function H of the user's language is,

[0086]

[0087] Among them, x i is the data information of the i-th feature value of the user's language, p is the complexity function of the language, λ i is the weight coefficient of the i-th eigenvalue of the user language, and n is the sample size.

[0088] In this embodiment, the complexity function p of the language is,

[0089]

[0090] Among them, x i It is the data information of the i-th feature value of the user's language.

[0091] In this embodiment, the lower the complexity, the higher the prediction certainty of the language model for the current word. Therefore, the training of the language model generally takes minimizing the complexity of the training sentence as the goal. To achieve this goal, the word frequency in the training sentence must be counted first to calculate the parameters of the language model. When the vocabulary is large and the training data is not sufficient, it will happen that the probability of some word sequences is very small or has never appeared.

[0092] In this embodiment, the weight coefficient λ of the i-th feature value of the user language is i The constraints are:

[0093]

[0094] Where n is the sample size.

[0095] In this embodiment, if Figure 4 As shown, the present invention provides a system for implementing any of the above-mentioned voice control methods applied to multiple servers, the system comprising: a data acquisition module, for acquiring data information of voice sequences of users of multiple groups of servers in real time based on voice receiving sensors of multiple servers, and performing preprocessing and feature extraction to obtain data information of feature vectors of voice sequences of users of multiple groups of servers;

[0096] A voice data optimization module, connected to the data acquisition module, is used to optimize the feature vector of the user's voice sequence using an improved White Whale optimization algorithm to obtain data information of the feature vector of the voice sequence of users of multiple groups of servers after optimization;

[0097] A language data feature value obtaining module, connected to the speech data optimization module, is used to construct a user's language model, analyze the feature vector of the user's speech sequence, and obtain data information of the feature value of the user's language;

[0098] The speech control output module is connected to the language data feature value obtaining module, and is used to construct a quality index function H of the user's language, characterize the quality of the user's language, obtain the quality characterization value of the user's language, set a preset threshold, and if the quality characterization value of the user's language is less than the preset threshold, the speech requirements are met, and the data information of the user's speech is output; if it is greater than the preset threshold, the speech requirements are not met and re-optimization is performed.

[0099] In this embodiment, the present invention provides a computer-readable storage medium, on which is stored a computer program programmed or configured to execute any one of the voice control methods applied to multiple servers.

[0100] Any reference to memory, storage, database or other media used in the embodiments provided in the present application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM) and memory bus dynamic RAM (RDRAM).

[0101] In summary, the present invention can not only accurately process the input voice and improve the quality of the output voice, but also process the voices received by multiple servers at the same time, thereby improving the operating efficiency of the system.

[0102] The above specific implementations do not constitute a limitation on the protection scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. A voice control method applied to multiple servers, characterized in that: The method comprises: U1. The user performs voice input, and the voice receiving sensors based on multiple servers obtain the data information of the voice sequences of users of multiple groups of servers in real time, and perform preprocessing and feature extraction to obtain the data information of the feature vectors of the voice sequences of users of multiple groups of servers; U2. Based on the data information of the feature vectors of the user's voice sequence of the multiple groups of servers, the feature vectors of the user's voice sequence are optimized using the improved White Whale optimization algorithm to obtain the data information of the feature vectors of the user's voice sequence of the optimized multiple groups of servers; U3. Based on the data information of the feature vectors of the user's voice sequence of the optimized multiple groups of servers, a user's language model is constructed, and the feature vectors of the user's voice sequence are analyzed to obtain data information of the feature values ​​of the user's language; U4. Based on the data information of the characteristic value of the user's language, construct the quality index function H of the user's language, characterize the quality of the user's language, obtain the quality characterization value of the user's language, set a preset threshold, if the quality characterization value of the user's language is less than the preset threshold, the voice requirements are met, and output the data information of the user's voice. If it is greater than the preset threshold, the voice requirements are not met, and return to step U2.

2. The voice control method applied to multiple servers according to claim 1, characterized in that: In step U4, the quality index function H of the user's language is, Among them, x i is the data information of the i-th feature value of the user's language, p is the complexity function of the language, λ i is the weight coefficient of the i-th eigenvalue of the user language, and n is the sample size.

3. The voice control method applied to multiple servers according to claim 2, characterized in that: The complexity function p of the language is, Among them, x i It is the data information of the i-th feature value of the user's language.

4. The voice control method applied to multiple servers according to claim 2, characterized in that: The weight coefficient λ of the i-th feature value of the user language i The constraints are: Where n is the sample size.

5. The voice control method applied to multiple servers according to claim 1, characterized in that: In step U2, the optimization of the feature vector of the user's speech sequence by using the improved White Whale optimization algorithm includes: U21. Based on the data information of the feature vectors of the speech sequences of the users of the multiple groups of servers, the beluga population is initialized, the population parameters and the maximum number of iterations T are determined, and the data information of the initialized beluga population is obtained; U22. Based on the data information of the initialized beluga population, establish the fitness function Q of the population individuals, Among them, y is the data information of the initialized beluga population, α1, α2 and α3 are the balance factors of the beluga population individuals, T is the maximum number of iterations, B0 is a random number in the interval (0,1), and the fitness values ​​of the population individuals are calculated to obtain the data information of the fitness values ​​of the population individuals; U23. Based on the data information of the fitness values ​​of the individuals in the population, establish a target optimization function W, Among them, z is the data information of the fitness value of the population individual, β1, β2 and β3 are random constant parameters in the interval (0,1), and the feature vector of the user's voice sequence is optimized to obtain the data information of the feature vector of the voice sequence of the users of the optimized multiple groups of servers.

6. The voice control method applied to multiple servers according to claim 5, characterized in that: The constraints of the balance factors α1, α2 and α3 of the beluga population individuals are:

7. The voice control method applied to multiple servers according to claim 1, characterized in that: In step U3, the construction of the user's language model and the analysis of the feature vector of the user's speech sequence include: U31. Based on the data information of the feature vectors of the voice sequences of the users of the optimized multiple groups of servers, a data set of feature vectors of the voice sequences of the users of the optimized multiple groups of servers is constructed; U32. Based on the data set of feature vectors of the user's voice sequence of the optimized multiple groups of servers, establish the feature value function R of the user's language, Wherein, r is the dataset of the feature vectors of the speech sequences of users of multiple groups of servers after optimization, and δ1, δ2 and δ3 are the weight coefficients of the feature vectors of the speech sequences; U33. Based on the characteristic value function R of the user's language, the characteristic vector of the user's speech sequence is analyzed to obtain data information of the characteristic value of the user's language.

8. The voice control method applied to multiple servers according to claim 7, characterized in that: The weight coefficients δ1, δ2 and δ3 of the feature vector of the speech sequence are, Wherein, r is a dataset of feature vectors of speech sequences of users of multiple groups of servers after optimization.

9. A system for implementing the voice control method applied to multiple servers according to any one of claims 1 to 8, characterized in that: The system comprises: A data acquisition module, used to acquire data information of speech sequences of users of multiple groups of servers in real time based on speech receiving sensors of multiple servers, and to perform preprocessing and feature extraction to obtain data information of feature vectors of speech sequences of users of multiple groups of servers; a speech data optimization module, connected to the data acquisition module, used to optimize the feature vectors of the user's speech sequences using an improved Beluga optimization algorithm to obtain data information of the optimized feature vectors of speech sequences of users of multiple groups of servers; A language data feature value obtaining module, connected to the speech data optimization module, is used to construct a user's language model, analyze the feature vector of the user's speech sequence, and obtain data information of the feature value of the user's language; The speech control output module is connected to the language data feature value obtaining module, and is used to construct a quality index function H of the user's language, characterize the quality of the user's language, obtain the quality characterization value of the user's language, set a preset threshold, and if the quality characterization value of the user's language is less than the preset threshold, the speech requirements are met, and the data information of the user's speech is output; if it is greater than the preset threshold, the speech requirements are not met and re-optimization is performed.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program programmed or configured to execute the voice control method applied to multiple servers as described in any one of claims 1 to 8.

Images