CN110347426B - Intelligent release APP platform system and method thereof - Google Patents

Intelligent release APP platform system and method thereof Download PDF

Info

Publication number
CN110347426B
CN110347426B CN201910735761.4A CN201910735761A CN110347426B CN 110347426 B CN110347426 B CN 110347426B CN 201910735761 A CN201910735761 A CN 201910735761A CN 110347426 B CN110347426 B CN 110347426B
Authority
CN
China
Prior art keywords
app
module
voice
keyword
current screen
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201910735761.4A
Other languages
Chinese (zh)
Other versions
CN110347426A (en
Inventor
杨明
孙煖
王欢
李翔
翁唯维
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Jilin Teachers Institute of Engineering and Technology
Original Assignee
Jilin Teachers Institute of Engineering and Technology
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Jilin Teachers Institute of Engineering and Technology filed Critical Jilin Teachers Institute of Engineering and Technology
Priority to CN201910735761.4A priority Critical patent/CN110347426B/en
Publication of CN110347426A publication Critical patent/CN110347426A/en
Application granted granted Critical
Publication of CN110347426B publication Critical patent/CN110347426B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90—Details of database functions independent of the retrieved data types
    • G06F16/95—Retrieval from the web
    • G06F16/958—Organisation or management of web site content, e.g. publishing, maintaining pages or automatic linking
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F8/00—Arrangements for software engineering
    • G06F8/60—Software deployment
    • G06F8/65—Updates
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F8/00—Arrangements for software engineering
    • G06F8/70—Software maintenance or management
    • G06F8/71—Version control; Configuration management
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F9/00—Arrangements for program control, e.g. control units
    • G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
    • G06F9/44—Arrangements for executing specific programs
    • G06F9/4401—Bootstrapping
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS OR SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING; SPEECH OR AUDIO CODING OR DECODING
    • G10L15/00—Speech recognition
    • G10L15/22—Procedures used during a speech recognition process, e.g. man-machine dialogue
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS OR SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING; SPEECH OR AUDIO CODING OR DECODING
    • G10L15/00—Speech recognition
    • G10L15/22—Procedures used during a speech recognition process, e.g. man-machine dialogue
    • G10L2015/223—Execution procedure of a spoken command

Abstract

The invention belongs to the technical field of intelligent APP popularization and discloses an intelligent APP release platform system and a method thereof; collecting voice instruction data by utilizing a sound sensor; preprocessing the collected voice command, and inputting search keywords by using an input keyboard; connecting the network card driver with the Internet for network communication; publishing the APP operation by utilizing a publishing program; searching APP information in the network according to the identified voice command or key word by using a search program; extracting the APP links by using an extraction program; the APP links are converted into two-dimension codes by utilizing a two-dimension code generator; downloading APP operation by using a downloading program; the operation of the APP is guided by a bootstrap program. The invention can solve the problems of higher error rate and waste of human resources caused by the existing manual operation, has good application prospect, has the advantages of high efficiency, difficult error occurrence and avoidance of waste of human resources, and has good application prospect.

Description

Intelligent release APP platform system and method thereof
Technical Field
The invention belongs to the technical field of intelligent APP popularization, and particularly relates to an intelligent APP release platform system and a method thereof.
Background
The intelligent APP mainly refers to software installed on the intelligent mobile phone, and improves the defects and individuation of an original system. The mobile phone can perfect the functions of the mobile phone, and provides a richer main means of using experience for voice instructions or keywords. In 2012, it is proposed that the model curves are crossed and presented with alpha to obtain names, and the APP promotion needs to look at the product type and the voice command or keyword positioning, and the priority needs to look at the budget, period and target of the APP promotion.
In a plurality of functional applications and game applications, the implantation of advertisements is the most basic mode, an advertiser performs advertisement implantation by implanting dynamic advertisement columns, and when a sound instruction or a keyword clicks the advertisement columns, the advertiser enters a website link, so that details of the advertiser can be known or the advertiser participates in an activity. However, the existing APP release needs to fill in a large amount of related information, is complicated, is easy to cause errors, and also causes waste of human resources; then manually checking whether the release application of the submitted APP meets the requirements, and manually allowing the APP to be released on an application release platform if the release application meets the requirements; meanwhile, for old voice instructions or keywords or voice instructions or keywords with insufficient learning capacity, APP operation is high in difficulty and easy to operate by mistake.
In summary, the problems of the prior art are:
(1) The existing APP release needs to fill in a large amount of related information, is complicated, is easy to cause errors, and also causes waste of human resources;
(2) Manually checking whether the release application of the submitted APP meets the requirements, and manually allowing the APP to be released on an application release platform if the release application meets the requirements, wherein the operation procedure is complicated;
(3) For voice instructions or keywords or voice instructions or keywords with insufficient learning capacity, APP operation is difficult and misoperation is easy;
(4) For the existing voice command recognition, the voice command recognition quality is poor because the voice signal is large in noise and other devices or surrounding environments can have larger influence on the collected voice data.
Disclosure of Invention
Aiming at the problems existing in the prior art, the invention provides an intelligent release APP platform system and a method thereof.
The invention is realized in such a way that the method for intelligently publishing the APP platform comprises the following steps:
searching APP information in the network according to the voice command or the keyword by using a search program; initializing a selected voice instruction or keyword information set and a candidate voice instruction or keyword information set in APP information retrieval in a network;
the search module initializes the following parameters, and initializes the selected voice instruction or keyword setCandidate voice instruction or keyword set A 0 ＝{1,2,3,L,L}，/>Representing an empty set, as the scheduling process proceeds, the selected voice command or keyword set and the elements of the candidate voice command or keyword set are continuously updated, A n And S is n Candidate and selected voice instructions or keyword sets at the end of the nth iteration, n=1, l, n, respectively T For the number of iterations, n=1 is initialized, and the voice command or keyword side can obtain channel matrix information by adopting a method of joint channel parameter estimation, namely, the search module transmits N T The method comprises the steps that a sub-signal stream is added with a data block consisting of a sound instruction or a training sequence with a known keyword side in front of each sub-signal stream, the sound instruction or the keyword realizes estimation of a channel state information matrix H between the sound instruction or the keyword and a retrieval module according to received signals and the known training data, and the sound instruction or the keyword sends the channel information matrix to the retrieval module;
channel information matrix H fed back by retrieval module to voice instruction or keyword k k Singular value decompositionWherein lambda is k,1 Singular values of the channel matrix representing the kth voice command or key, reflecting the transmission gain of the voice command or key channel, +.>Representing dimensions 1× (N T Zero vector of-1), ∈>And->Respectively by non-zero singular values lambda k,1 Right singular value vector corresponding to zero singular value is constructed because rank (H k ) =1, so->v i,1 Is V (V) i Wherein rank (·) represents the rank of the matrix;
the retrieval module constructs an intermediate matrix according to the decomposed matrixAnd-> AndWherein diag (·) represents the diagonalization operation;
the search module constructs a correlation matrix R, wherein R is an L multiplied by L square matrix, and the elements of the ith row and the jth column areWherein, I·| represents modulo operation, r i,j The degree of correlation between the reflected voice command or key i and j;
selecting n-1 columns corresponding to the n-1 voice instructions or keywords to be scheduled from R to form a matrixThe remaining parts are arranged in ascending order to obtain matrix +.>I.e. < ->
Calculating correlation factors for voice commands or keywords according toI.e. pair R n The first xi elements of each row are respectively summed and the reciprocal is taken;
equivalent to the simplification of
Obtaining a column vector psi n ＝[ψ 1,n L ψ L,n ] Η Wherein A is l Representing a potential, possibly subsequently selected set of voice instructions or keywords, the card (·) representing the number of elements in the set,is the pair +.>Because the selected voice instruction or keyword set S is not finalized before the end of the scheduling process, ψ k The calculation cannot be accurately performed;
the nth voice command or keyword is selected according to the following formula:
s n a label representing the selected voice command or keyword,is the scheduling weight of the voice instruction or key k,/->Is the average correlation factor of the tone command or key k at the end of the last transmission period, update S n ＝S n-1 U{s n }，A n ＝A n-1 -{s n }，n＝n+1；
If N is less than N T Returning to obtain a matrixA step of; otherwise, the dispatching is completed according to the dispatched voice instruction or key wordThe actual interference experienced calculates the correlation factor ψ k The method comprises the steps of carrying out a first treatment on the surface of the If voice instruction or key word is not scheduled +>ψ k =0, and updates the pitch instruction or key k, k e {1, l } average correlation factor as follows, for calculating the pitch instruction or key scheduling weight in the next transmission cycle,
wherein delta c After the scheduling is completed, the retrieving module notifies the activate voice command or key word and performs downlink data communication, and repeatedly executes the above steps in the overhead time slot stage of the next transmission period (t+1) to retrieve APP information in the network.
Further, the method for intelligently publishing the APP platform specifically comprises the following steps:
step one, voice instruction data are collected by a voice collection module through a voice sensor; the method comprises the steps of inputting search keywords by using an input keyboard through a data input module;
step two, the main control module is connected with the Internet through a network communication module by utilizing a network card driver to carry out network communication;
step three, publishing APP operation by utilizing a publishing program through a publishing module; searching APP information in the network by using a search program according to the voice instruction or the keyword through a search module;
step four, extracting the APP links by using an extraction program through a link acquisition module; the APP is linked and converted into a two-dimensional code by a two-dimensional code generating module through a two-dimensional code generator;
step five, downloading APP operation by using a downloading program through a downloading module; guiding the operation of the APP by using a guiding program through an operation guiding module;
and step six, displaying voice instructions, search keywords, APP information and two-dimensional code data by using a display module.
Further, the publishing module publishing method comprises the following steps:
(1) After development is completed, the APP is issued and configured, and the first packing operation is executed, so that a data packet is obtained;
(2) An application master submits a request for releasing the APP after the first package on an application release platform;
(3) The application release platform carries out automatic auditing on the APP, and the auditing contents comprise: the method comprises the steps that identity and real name authentication of an APP master, MUA number of the APP on the Internet and content in the APP are carried out, and when the content meets requirements, an application release platform judges that the APP passes auditing;
(4) After passing the verification, the APP is packaged for the second time, and the verification passing label is packaged into the APP and allowed to be released on an application release platform.
Further, the APP is configured on an APP development platform, the APP development platform issues update information aiming at a specific type of APP, and the APP marked with a 'pass through' label in the APP development platform is detected, and update prompt information is issued only for the APP;
after the first packing operation is finished, the APP is uploaded to a file server, only the APP host is allowed to download the APP, and the viewing or editing operation is executed;
the APP content downloaded through the application release platform is the content passing the audit, and the APP content downloaded through the APP development platform is the content edited by the application master.
Further, the operation guiding module guiding method includes:
1) Downloading an APP through a downloading program, and after the APP is started, sampling a current screen if target guide software is already operated;
2) Extracting a characteristic image from the sampled picture, and searching a guiding scheme corresponding to the characteristic image;
3) Displaying on the current screen according to the guiding scheme to finish guiding;
further, the displaying on the current screen according to the guiding scheme to complete guiding includes:
(1) Displaying selection information corresponding to the guiding scheme on the current screen;
(2) Acquiring a voice command or a keyword input selection command;
(3) Invoking a guide picture set corresponding to the selection instruction;
(4) Sequentially displaying each guide picture in the guide picture set on the current screen;
further, the displaying each guidance picture in the guidance picture set on the current screen in turn includes:
(1) If the display picture on the current screen changes, extracting the characteristics of the changed current screen;
(2) And if the extracted characteristics of the changed current screen accord with the rule corresponding to the selection instruction, displaying a guide picture corresponding to the extracted characteristics of the changed current screen on the current screen.
Further, the feature extraction of the changed current screen includes:
according to the guiding step of the current screen before the change, calculating the guiding step after the change;
searching a characteristic image area corresponding to the changed guiding step;
and extracting the characteristics in the characteristic image area of the current screen to determine the characteristic identification.
Further, the displaying on the current screen according to the guiding scheme to complete guiding further includes:
and acquiring voice data matched with the target guide picture displayed on the current screen.
Further, the displaying on the current screen according to the guiding scheme to complete guiding further includes: and playing the voice data matched with the target guide picture when the target guide picture is displayed on the current screen.
Another object of the present invention is to provide an intelligent distribution APP platform, including:
(1) The voice acquisition module is connected with the main control module and used for acquiring voice instruction data through the voice sensor;
(2) The data input module is connected with the main control module and is used for inputting search keywords through an input keyboard;
(3) The main control module is connected with the voice acquisition module, the data input module, the network communication module, the publishing module, the retrieval module, the link acquisition module, the two-dimension code generation module, the downloading module, the operation guiding module and the display module and used for controlling each module to work normally through the singlechip;
(4) The network communication module is connected with the main control module and is used for carrying out network communication by connecting the network card driver with the Internet;
(5) The release module is connected with the main control module and used for releasing APP operation through a release program;
(6) The retrieval module is connected with the main control module and is used for retrieving APP information in the network according to the voice instruction or the keyword through a retrieval program;
(7) The link acquisition module is connected with the main control module and is used for extracting an APP link through an extraction program;
(8) The two-dimensional code generation module is connected with the main control module and used for converting the APP link into a two-dimensional code through the two-dimensional code generator;
(9) The downloading module is connected with the main control module and used for downloading APP operation through a downloading program;
(10) The operation guiding module is connected with the main control module and is used for guiding the operation of the APP through a guiding program;
(11) The display module is connected with the main control module and used for displaying voice instructions, search keywords, APP information and two-dimensional code data through the display.
The invention has the advantages and positive effects that:
(1) The invention can realize the automatic checking whether the release application of the APP meets the requirement or not through the release module, and automatically allows the APP to be released on the application release platform after the checking is passed, thereby being capable of well solving the problems of higher error rate and waste of human resources caused by the prior manual operation;
(2) The method has good application prospect, has the advantages of high efficiency, difficult error occurrence and avoiding the waste of human resources, and has good application prospect;
(3) The software executing the guiding function can automatically identify the content of the current screen before the old people use the software by operating the guiding module, and the corresponding guiding scheme is matched for guiding, so that the impression of voice instructions or keywords on the use rules is enhanced, and the misoperation probability of the software is reduced.
The invention eliminates the influence on the quality of the voice signal by preprocessing the collected voice signal and eliminating the factors such as aliasing, higher harmonic distortion, high frequency and the like caused by the human voice organ and the equipment for collecting the voice signal. The method ensures that the signals obtained by the subsequent voice processing are more uniform and smoother as far as possible, provides high-quality parameters for signal parameter extraction, and improves the voice processing quality. The pre-emphasis of the voice signal can emphasize the high frequency part of the voice, remove the influence of the radiation of the lips and increase the high frequency resolution of the voice.
The invention can smooth the frequency spectrum and eliminate the function of harmonic wave by selecting the triangular band-pass filter when recognizing the voice signal, and highlight the formants of the original voice. The tone or pitch of a segment of sound is therefore not reflected within the MFCC parameters, i.e. the MFCC is used as acoustic feature, and is not affected by the difference in tone of the input sound. The amount of computation can also be reduced.
According to the method, APP information in a network is searched by utilizing a search program according to voice instructions or keywords; initializing a selected voice instruction or keyword information set and a candidate voice instruction or keyword information set in APP information retrieval in a network; the search module initializes the following parameters, and initializes the selected voice instruction or keyword setCandidate voice instruction or keyword set A 0 ＝{1,2,3,L,L}，/>Representing an empty set, as the scheduling process proceeds, the selected voice command or keyword set and the elements of the candidate voice command or keyword set are continuously updated, A n And S is n Candidate and selected voice instructions or keyword sets at the end of the nth iteration, n=1, l, n, respectively T For the number of iterations, n=1 is initialized, and the voice command or keyword side can obtain channel matrix information by adopting a method of joint channel parameter estimation, namely, the search module transmits N T The method comprises the steps that a sub-signal stream is added with a data block consisting of a sound instruction or a training sequence with a known keyword side in front of each sub-signal stream, the sound instruction or the keyword realizes estimation of a channel state information matrix H between the sound instruction or the keyword and a retrieval module according to received signals and the known training data, and the sound instruction or the keyword sends the channel information matrix to the retrieval module; finally, the searching module informs the activating voice instruction or key word and performs downlink data communication, and searches the network in the overhead time slot stage of the next transmission period (t+1)APP information in the complex.
According to the channel matrix H of the obtained voice command or keyword, singular value decomposition is carried out, each received interference is estimated by constructing an intermediate matrix and a correlation matrix, a weight is given to each received interference, the voice command or keyword is selected by taking the weighted correlation maximum as a criterion, a group of voice commands or keywords with small interference among each other are reasonably selected, and the system and rate rationality are realized.
Drawings
Fig. 1 is a flowchart of a method for intelligently publishing an APP platform provided by an embodiment of the present invention.
FIG. 2 is a block diagram of an intelligent issuing APP platform provided by an embodiment of the invention;
in the figure: 1. a voice acquisition module; 2. a data input module; 3. a main control module; 4. a network communication module; 5. a release module; 6. a retrieval module; 7. a link acquisition module; 8. a two-dimensional code generating module; 9. downloading a module; 10. an operation guide module; 11. and a display module.
Detailed Description
For a further understanding of the invention, its features and advantages, reference is now made to the following examples, which are illustrated in the accompanying drawings.
The existing APP release needs to fill in a large amount of related information, is complicated, is easy to cause errors, and also causes waste of human resources; then manually checking whether the release application of the submitted APP meets the requirements, and manually allowing the APP to be released on an application release platform if the release application meets the requirements; meanwhile, for old voice instructions or keywords or voice instructions or keywords with insufficient learning capacity, APP operation is high in difficulty and easy to operate by mistake. In the existing voice command recognition, the voice signal is large in noise, and other devices or surrounding environments can have larger influence on collected voice data, so that the voice command recognition quality is poor.
In order to solve the above problems, the present invention will be described in detail with reference to specific embodiments.
As shown in fig. 1, the method for intelligently publishing an APP platform provided by the embodiment of the invention includes the following steps:
s101, voice instruction data are collected by utilizing a sound sensor; preprocessing the collected voice command, and inputting search keywords by using an input keyboard.
S102, connecting the network card driver to the Internet for network communication.
S103, publishing the APP operation by utilizing a publishing program; and searching APP information in the network according to the identified voice command or key words by using a search program.
S104, extracting an APP link by using an extraction program; and (5) utilizing a two-dimension code generator to generate a two-dimension code by performing APP link conversion.
S105, downloading APP operation by using a downloading program; the operation of the APP is guided by a bootstrap program.
And S106, displaying voice instructions, search keywords, APP information and two-dimensional code data by using a display.
In step S101, the voice preprocessing provided by the embodiment of the present invention includes:
(1) Firstly, sampling an analog voice signal s (T) in a sampling period T, discretizing the analog voice signal s (n), and determining the period according to the bandwidth (Nyquist sampling theorem) of the analog voice signal to avoid frequency domain aliasing distortion of the signal; the frequency range of the speech signal is usually 300-3400 Hz, and the sampling rate is usually 8 KHZ.
(2) Pre-emphasis is performed on the input digital speech signal, typically by a transfer function of:
H(z)＝1-αz -1
wherein a is a pre-emphasis coefficient, 0.9 < a <1.0.
Let the voice sampling value at time n be x (n), the result after pre-emphasis treatment is:
y(n)＝x(n)-αx(n-1)
a＝0.98。
(3) Carrying out windowing and framing treatment, and framing a voice signal by adopting a movable limited-length window weighting method; the number of frames per second is generally 33 to 100 frames, as the case may be.
Framing by adopting an overlapping segmentation method; the overlapping portion of the preceding frame and the following frame is called frame shift, and the ratio of frame shift to frame length is generally 0 to 1/2.
With Hamming (Hamming) windows, the window function includes:
(4) After the window function is determined, the framing of the speech signal is performed, in effect, by performing some transform or operation on each frame.
Let the transform or operation be denoted by T [ ], x (n) be the input speech signal, w (n) be the window sequence, h (n) be the filter associated with w (n), the processed output of each frame can be expressed as:
(5) Extracting the characteristics of the signals input by each frame to obtain a characteristic vector X i The method comprises the steps of carrying out a first treatment on the surface of the The dimension is K; the feature vector is encoded by a VQ encoder, and the channel transmission and the processing are carried out; and (5) entering a VQ decoder to obtain a feature vector for further processing.
(6) When the voice recognition is carried out, the similarity between the vector sequence obtained by carrying out feature extraction on the input voice signal and the reference vector is calculated, the word (word) corresponding to the codebook with the smallest total average distortion error is selected as a recognition result according to the optimal matching criterion, and then the recognized content is output, namely the understanding of the computer on the natural voice.
In step S101, searching APP information in the network according to the voice command or the keyword by using a search program; in APP information retrieval in a network, a selected voice command or keyword information set and a candidate voice command or keyword information set are initialized.
The search module initializes the following parameters, and initializes the selected voice instruction or keyword setCandidate voice instruction or keyword set A 0 ＝{1,2,3,L,L}，/>Representing an empty set, as the scheduling process proceeds, the selected voice command or keyword set and the elements of the candidate voice command or keyword set are continuously updated, A n And S is n Candidate and selected voice instructions or keyword sets at the end of the nth iteration, n=1, l, n, respectively T For the number of iterations, n=1 is initialized, and the voice command or keyword side can obtain channel matrix information by adopting a method of joint channel parameter estimation, namely, the search module transmits N T And the sub-signal flows are added with data blocks consisting of a sound instruction or a key word side known training sequence before each sub-signal flow, the sound instruction or the key word realizes the estimation of a channel state information matrix H between the sound instruction or the key word and the retrieval module according to the received signals and the known training data, and the sound instruction or the key word sends the channel information matrix to the retrieval module.
Channel information matrix H fed back by retrieval module to voice instruction or keyword k k Singular value decompositionWherein lambda is k,1 Singular values of the channel matrix representing the kth voice command or key, reflecting the transmission gain of the voice command or key channel, +.>Representing dimensions 1× (N T The zero vector of-1),and->Respectively by non-zero singular values lambda k,1 Right singular value vector corresponding to zero singular value is constructed because rank (H k ) =1, so->v i,1 Is V (V) i Wherein rank (·) represents the rank of the matrix.
The retrieval module constructs an intermediate matrix according to the decomposed matrixAnd-> AndWherein diag (·) represents the diagonalization operation.
The search module constructs a correlation matrix R, wherein R is an L multiplied by L square matrix, and the elements of the ith row and the jth column areWherein, I·| represents modulo operation, r i,j The degree of correlation between the reflected voice command or key i and j.
Selecting n-1 columns corresponding to the n-1 voice instructions or keywords to be scheduled from R to form a matrixThe remaining parts are arranged in ascending order to obtain matrix +.>I.e. < ->
Calculating correlation factors for voice commands or keywords according toI.e. pair R n The first ζ elements of each row are summed and inverted, respectively.
Equivalent to the simplification of
Obtaining a column vector psi n ＝[ψ 1,n L ψ L,n ] Η Wherein A is l Representing a potential, possibly subsequently selected set of voice instructions or keywords, the card (·) representing the number of elements in the set,is the pair +.>Because the selected voice instruction or keyword set S is not finalized before the end of the scheduling process, ψ k And cannot be accurately calculated.
The nth voice command or keyword is selected according to the following formula:
s n a label representing the selected voice command or keyword,is the scheduling weight of the voice instruction or key k,/->Is the average correlation factor of the tone command or key k at the end of the last transmission period, update S n ＝S n-1 U{s n }，A n ＝A n-1 -{s n }，n＝n+1。
If N is less than N T Returning to obtain a matrixA step of; otherwise, the dispatching is completed according to the dispatched voice instruction or key wordThe actual interference experienced calculates the correlation factor ψ k The method comprises the steps of carrying out a first treatment on the surface of the If voice instruction or key word is not scheduled +>ψ k =0, and updates the pitch instruction or key k, k e {1, l } average correlation factor as follows, for calculating the pitch instruction or key scheduling weight in the next transmission cycle,
wherein delta c After the scheduling is completed, the retrieving module notifies the activate voice command or key word and performs downlink data communication, and repeatedly executes the above steps in the overhead time slot stage of the next transmission period (t+1) to retrieve APP information in the network.
In step S103, the voice command recognition provided in the embodiment of the present invention includes:
(1) Performing fast Fourier transform on each frame of signals subjected to framing and windowing to obtain a frequency spectrum of each frame, and performing modular squaring on the frequency spectrum of the voice signal to obtain a power spectrum of the voice signal; let DFT of the speech signal be:
where x [ N ] is the input speech signal and N represents the number of points of the Fourier transform.
(2) The frequency of the Fourier transform output is corresponding to the mel scale; a mel is a unit of pitch where sounds that are perceived equidistant in pitch can be separated by the same number of mel numbers; the correspondence between frequency (in Hz) and mel scale is linear below 1000Hz and logarithmic above 1000Hz, the calculation formula of which comprises:
defining a filter group with M filters, wherein the adopted filters are triangular filters, the center frequency is f (M), m=1, 2, …, and M is usually 22-26 (the number of the filters is similar to the number of critical bands); the interval between f (m) decreases as the value of m decreases and widens as the value of m increases, the frequency response of each triangular filter being:
after obtaining mel frequency spectrum, calculating the logarithmic energy output by each filter bank; the logarithmic energy output by each filter is:
(3) And (3) cepstrum: discrete Cosine Transform (DCT), i.e. taking the logarithm of the amplitude value of the standard amplitude spectrum, and then visualizing the logarithmic spectrum to look like a sound waveform;
using the logarithmic energy of the filter, the cepstral coefficients can be obtained from the discrete cosine transform:
where L refers to the MFCC order, typically 12 can represent the acoustic feature; m refers to the number of triangular filters.
(4) Energy and look-up, the energy of a frame is defined as the sum of squares of sample points of a frame, and for a windowed signal x, the energy from sample point t1 to sample point t2 is:
calculating the difference value of 13 characteristics of each frame before and after the current frame:
(5) And carrying out matching recognition on the input MFCC parameter vector.
As shown in fig. 2, the method for intelligently publishing an APP platform provided by the embodiment of the invention includes: the system comprises a voice acquisition module 1, a data input module 2, a main control module 3, a network communication module 4, a release module 5, a retrieval module 6, a link acquisition module 7, a two-dimension code generation module 8, a downloading module 9, an operation guiding module 10 and a display module 11.
The voice acquisition module 1 is connected with the main control module 3 and is used for acquiring voice instruction data through the voice sensor.
And the data input module 2 is connected with the main control module 3 and is used for inputting search keywords through an input keyboard.
The main control module 3 is connected with the voice acquisition module 1, the data input module 2, the network communication module 4, the publishing module 5, the retrieval module 6, the link acquisition module 7, the two-dimensional code generation module 8, the downloading module 9, the operation guiding module 10 and the display module 11 and is used for preprocessing and recognizing voice instructions and controlling the normal work of each module through the singlechip.
And the network communication module 4 is connected with the main control module 3 and is used for carrying out network communication by connecting the network card driver with the Internet.
And the release module 5 is connected with the main control module 3 and used for releasing APP operation through a release program.
And the retrieval module 6 is connected with the main control module 3 and is used for retrieving APP information in the network according to the voice instruction or the keyword through a retrieval program.
And the link acquisition module 7 is connected with the main control module 3 and is used for extracting the APP link through an extraction program.
And the two-dimension code generation module 8 is connected with the main control module 3 and is used for generating the two-dimension code by linking and converting the APP through the two-dimension code generator.
And the downloading module 9 is connected with the main control module 3 and used for downloading APP operation through a downloading program.
The operation guiding module 10 is connected with the main control module 3 and is used for guiding the operation of the APP through a guiding program;
the display module 11 is connected with the main control module 3 and is used for displaying voice instructions, search keywords, APP information and two-dimensional code data through a display.
The publishing module 5 publishing method provided by the embodiment of the invention comprises the following steps:
(1) After development is completed, the APP is issued and configured, and the first packing operation is executed to obtain a data packet.
(2) The application master submits a request for releasing the APP after the first package on an application release platform.
(3) The application release platform carries out automatic auditing on the APP, and the auditing contents comprise: and after the content meets the requirements, the application release platform judges that the APP passes the auditing.
(4) After passing the verification, the APP is packaged for the second time, and the verification passing label is packaged into the APP and allowed to be released on an application release platform.
The APP provided by the embodiment of the invention comprises the following steps:
after configuration is completed on an APP development platform, the APP development platform issues update information aiming at a specific type of APP, and the APP marked with a 'verification passing' label in the APP development platform is detected, and update prompt information is issued only for the APP; after the first packing operation is finished, the APP is uploaded to a file server, only the APP host is allowed to download the APP, and the viewing or editing operation is executed; the APP content downloaded through the application release platform is the content passing the audit, and the APP content downloaded through the APP development platform is the content edited by the application master.
The method for guiding the operation guiding module 10 provided by the embodiment of the invention comprises the following steps:
(1) Downloading an APP through a downloading program, and after the APP is started, sampling a current screen if target guide software is already operated;
(2) Extracting a characteristic image from the sampled picture, and searching a guiding scheme corresponding to the characteristic image;
(3) Displaying selection information corresponding to the guiding scheme on a current screen; acquiring a voice command or a keyword input selection command; invoking a guide picture set corresponding to the selection instruction; if the display picture on the current screen changes, extracting the characteristics of the changed current screen; if the extracted characteristics of the changed current screen accord with the rule corresponding to the selection instruction, displaying a guide picture corresponding to the extracted characteristics of the changed current screen on the current screen; acquiring voice data matched with a target guide picture displayed on a current screen; and playing the voice data matched with the target guide picture when the target guide picture is displayed on the current screen in the executing step.
In step (3), the feature extraction for the changed current screen provided by the embodiment of the invention comprises the following steps:
according to the guiding step of the current screen before the change, calculating the guiding step after the change; searching a characteristic image area corresponding to the changed guiding step; and extracting the characteristics in the characteristic image area of the current screen to determine the characteristic identification.
The foregoing description is only of the preferred embodiments of the present invention, and is not intended to limit the invention in any way, but any simple modification, equivalent variation and modification of the above embodiments according to the technical principles of the present invention are within the scope of the technical solutions of the present invention.

Claims (10)

1. The method for intelligently publishing the APP platform is characterized by comprising the following steps of:
searching APP information in the network according to the voice command or the keyword by using a search program; initializing a selected voice instruction or keyword information set and a candidate voice instruction or keyword information set in APP information retrieval in a network;
the search module initializes the following parameters, and initializes the selected voice instruction or keyword setCandidate voice instruction or keyword set A 0 ＝{1,2,3,…,L}，/>Representing an empty set, as the scheduling process proceeds, the selected voice command or keyword set and the elements of the candidate voice command or keyword set are continuously updated, A n And S is n Candidate and selected voice instructions or keyword sets at the end of the nth iteration, n=1, …, N, respectively T For the number of iterations, initializing n=1, and obtaining channel matrix information by adopting a joint channel parameter estimation method at the voice instruction or key word side, namely transmitting N by a retrieval module T The method comprises the steps that a sub-signal stream is added with a data block consisting of a sound instruction or a training sequence with a known keyword side in front of each sub-signal stream, the sound instruction or the keyword realizes estimation of a channel state information matrix H between the sound instruction or the keyword and a retrieval module according to received signals and the known training data, and the sound instruction or the keyword sends the channel information matrix to the retrieval module;
channel information matrix H fed back by retrieval module to voice instruction or keyword k k Singular value decompositionWherein lambda is k,1 Singular values of the channel matrix representing the kth voice command or key, reflecting the transmission gain of the voice command or key channel, +.>Representing dimensions 1× (N T The zero vector of-1),and->Respectively by non-zero singular values lambda k,1 Right singular value vector corresponding to zero singular value is constructed because rank (H k ) =1, so->v i,1 Is V (V) i Wherein rank (·) represents the rank of the matrix; parameter rank (H) k ) Is defined as a channel information matrix H k Rank of (c);
the retrieval module constructs an intermediate matrix according to the decomposed matrixAnd-> AndWherein (1)>v L，1 V is L Is the first column vector of (a); lambda (lambda) L，1 Singular values of the channel matrix representing the L-th voice command or key, reflecting the transmission gain of the voice command or key channel, diag (·) representing the diagonalization operation;
the searching module constructs a correlation matrix R, wherein R is an L multiplied by L square matrix, and the ith row and the jth columnThe elements areWherein, I·| represents modulo operation, r i,j The degree of correlation between the reflected voice command or key i and j; />v i,1 Is V (V) i Is the first column vector of (a); /> Is V (V) j Is a first row vector of (a);
selecting n-1 columns corresponding to the n-1 voice instructions or keywords to be scheduled from R to form a matrixThe remaining parts are arranged in ascending order to obtain matrix +.>I.e. < ->
Calculating correlation factors for voice commands or keywords according toI.e. pair R n The first xi elements of each row are respectively summed and the reciprocal is taken;
equivalent to simplifying the formula:wherein; r is (r) k，j The degree of correlation between the reflected voice command or key words k and j;
obtaining a column vector psi n ＝[ψ 1,n … ψ L,n ] Η Wherein A is l Representing a potential, possibly subsequently selected set of voice instructions or keywords, the card (·) representing the number of elements in the set,is the pair +.>Because the selected voice instruction or keyword set S is not finalized before the end of the scheduling process, ψ k The calculation cannot be accurately performed;
the nth voice command or keyword is selected according to the following formula:
s n a label representing the selected voice command or keyword,is the scheduling weight of the voice instruction or key k,/->Is the average correlation factor of the tone command or key k at the end of the last transmission period, update S n ＝S n-1 ∪{s n }，A n ＝A n-1 -{s n }，n＝n+1；
If N is less than N T Returning to obtain a matrixA step of; otherwise, the scheduling is completed according to the scheduled voice instruction or key word +>The actual interference experienced calculates the correlation factor ψ k The method comprises the steps of carrying out a first treatment on the surface of the If voice instruction or key word is not scheduled +>ψ k =0, and updates the pitch instruction or key k, k e {1, …, L } average correlation factor as follows, for calculating the pitch instruction or key scheduling weight in the next transmission cycle,
wherein delta c After the scheduling is completed, the retrieving module notifies the activate voice command or key word and performs downlink data communication, and repeatedly executes the above steps in the overhead time slot stage of the next transmission period (t+1) to retrieve APP information in the network.
2. The method for intelligently publishing an APP platform according to claim 1, wherein the method for intelligently publishing an APP platform specifically comprises the following steps:
step one, voice instruction data are collected by a voice collection module through a voice sensor; the method comprises the steps of inputting search keywords by using an input keyboard through a data input module;
step two, the main control module is connected with the Internet through a network communication module by utilizing a network card driver to carry out network communication;
step three, publishing APP operation by utilizing a publishing program through a publishing module; searching APP information in the network by using a search program according to the voice instruction or the keyword through a search module;
step four, extracting the APP links by using an extraction program through a link acquisition module; the APP is linked and converted into a two-dimensional code by a two-dimensional code generating module through a two-dimensional code generator;
step five, downloading APP operation by using a downloading program through a downloading module; guiding the operation of the APP by using a guiding program through an operation guiding module;
and step six, displaying voice instructions, search keywords, APP information and two-dimensional code data by using a display module.
3. The method for intelligently publishing an APP platform according to claim 2, wherein the publishing module publishing method comprises:
(1) After development is completed, the APP is issued and configured, and the first packing operation is executed, so that a data packet is obtained;
(2) An application master submits a request for releasing the APP after the first package on an application release platform;
(3) The application release platform carries out automatic auditing on the APP, and the auditing contents comprise: the method comprises the steps that identity and real name authentication of an APP master, MUA number of the APP on the Internet and content in the APP are carried out, and when the content meets requirements, an application release platform judges that the APP passes auditing;
(4) After passing the verification, the APP is packaged for the second time, and the verification passing label is packaged into the APP and allowed to be released on an application release platform.
4. The method for intelligently publishing an APP platform according to claim 3, wherein the APP application is configured on an APP development platform, the APP development platform publishes update information aiming at the APP application, and detects the APP application marked with a 'pass' tag in the APP development platform, and only publishes update prompt information for the APP application; after the first packing operation is finished, the APP is uploaded to a file server, only the APP host is allowed to download the APP, and the viewing or editing operation is executed; the APP content downloaded through the application release platform is the content passing the audit, and the APP content downloaded through the APP development platform is the content edited by the application master.
5. The method for intelligently publishing an APP platform according to claim 2, wherein the operation guiding module guiding method comprises:
(1) Downloading an APP through a downloading program, and after the APP is started, sampling a current screen if target guide software is already operated;
(2) Extracting a characteristic image from the sampled picture, and searching a guiding scheme corresponding to the characteristic image;
(3) And displaying on the current screen according to the guiding scheme to finish guiding.
6. The method of intelligently publishing an APP platform according to claim 5, wherein said displaying on said current screen according to said booting scheme to complete booting comprises:
(1) Displaying selection information corresponding to the guiding scheme on the current screen;
(2) Acquiring a voice command or a keyword input selection command;
(3) Invoking a guide picture set corresponding to the selection instruction;
(4) And displaying each guide picture in the guide picture set on the current screen in turn.
7. The method of intelligent distribution APP platform of claim 6, wherein displaying each guidance picture in the set of guidance pictures on the current screen in turn comprises:
(1) If the display picture on the current screen changes, extracting the characteristics of the changed current screen;
(2) If the extracted characteristics of the changed current screen accord with the rule corresponding to the selection instruction, displaying a guide picture corresponding to the extracted characteristics of the changed current screen on the current screen;
the feature extraction of the changed current screen comprises the following steps:
(1) According to the guiding step of the current screen before the change, calculating the guiding step after the change;
(2) Searching a characteristic image area corresponding to the changed guiding step;
(3) And extracting the characteristics in the characteristic image area of the current screen to determine the characteristic identification.
8. The method of intelligently publishing an APP platform according to claim 5, wherein said displaying on said current screen according to said booting scheme to complete booting further comprises:
and acquiring voice data matched with the target guide picture displayed on the current screen.
9. The method of intelligently publishing an APP platform of claim 5, wherein displaying on a current screen according to a boot scheme to complete a boot further comprises: and playing the voice data matched with the target guide picture when the target guide picture is displayed on the current screen.
10. An intelligent distribution APP platform for implementing the method for intelligently distributing APP platforms of claim 1, said intelligent distribution APP platform comprising:
the voice acquisition module is connected with the main control module and used for acquiring voice instruction data through the voice sensor;
the data input module is connected with the main control module and is used for inputting search keywords through an input keyboard;
the main control module is connected with the voice acquisition module, the data input module, the network communication module, the publishing module, the retrieval module, the link acquisition module, the two-dimension code generation module, the downloading module, the operation guiding module and the display module and used for controlling each module to work normally through the singlechip;
the network communication module is connected with the main control module and is used for carrying out network communication by connecting the network card driver with the Internet;
the release module is connected with the main control module and used for releasing APP operation through a release program;
the retrieval module is connected with the main control module and is used for retrieving APP information in the network according to the voice instruction or the keyword through a retrieval program;
the link acquisition module is connected with the main control module and is used for extracting an APP link through an extraction program;
the two-dimensional code generation module is connected with the main control module and used for converting the APP link into a two-dimensional code through the two-dimensional code generator;
the downloading module is connected with the main control module and used for downloading APP operation through a downloading program;
the operation guiding module is connected with the main control module and is used for guiding the operation of the APP through a guiding program;
the display module is connected with the main control module and used for displaying voice instructions, search keywords, APP information and two-dimensional code data through the display.
CN201910735761.4A 2019-08-09 2019-08-09 Intelligent release APP platform system and method thereof Active CN110347426B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910735761.4A CN110347426B (en) 2019-08-09 2019-08-09 Intelligent release APP platform system and method thereof

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910735761.4A CN110347426B (en) 2019-08-09 2019-08-09 Intelligent release APP platform system and method thereof

Publications (2)

Publication Number Publication Date
CN110347426A CN110347426A (en) 2019-10-18
CN110347426B true CN110347426B (en) 2024-01-26

Family

ID=68184505

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910735761.4A Active CN110347426B (en) 2019-08-09 2019-08-09 Intelligent release APP platform system and method thereof

Country Status (1)

Country Link
CN (1) CN110347426B (en)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112394982B (en) * 2021-01-21 2021-04-13 腾讯科技（深圳）有限公司 Method, device, medium and electronic equipment for generating voice recognition system
CN117608554A (en) * 2023-11-24 2024-02-27 上海浦东发展银行股份有限公司 Visual voice flow one-key release method, system and medium

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20050080625A1 (en) * 1999-11-12 2005-04-14 Bennett Ian M. Distributed real time speech recognition system
CN107861753A (en) * 2017-06-26 2018-03-30 平安普惠企业管理有限公司 APP generations index, search method and system and readable storage medium storing program for executing
CN107968819A (en) * 2017-11-20 2018-04-27 扬州通彩网络科技有限公司 A kind of new task management system and its operating procedure based on APP
CN109615436A (en) * 2018-12-17 2019-04-12 广州市统云网络科技有限公司 A kind of intelligent advertisement distribution platform, system and method based on internet screen

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20050080625A1 (en) * 1999-11-12 2005-04-14 Bennett Ian M. Distributed real time speech recognition system
CN107861753A (en) * 2017-06-26 2018-03-30 平安普惠企业管理有限公司 APP generations index, search method and system and readable storage medium storing program for executing
CN107968819A (en) * 2017-11-20 2018-04-27 扬州通彩网络科技有限公司 A kind of new task management system and its operating procedure based on APP
CN109615436A (en) * 2018-12-17 2019-04-12 广州市统云网络科技有限公司 A kind of intelligent advertisement distribution platform, system and method based on internet screen

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
基于AndroidAndroidAndroidAndroidAndroidAndroidAndroid平台的校园APPAPPAPP的设计与实现的设计与实现;李笑满;《中国优秀博硕士学位论文全文数据库(硕士)工程科技Ⅱ辑》;全文 *

Also Published As

Publication number Publication date
CN110347426A (en) 2019-10-18

Similar Documents

Publication Publication Date Title
CN107610709B (en) Method and system for training voiceprint recognition model
CN112802448B (en) Speech synthesis method and system for generating new tone
CN110853670B (en) Music-driven dance generation method
CN110415687A (en) Method of speech processing, device, medium, electronic equipment
CN107077842A (en) System and method for phonetic transcription
CN112071330B (en) Audio data processing method and device and computer readable storage medium
Kelly et al. Deep neural network based forensic automatic speaker recognition in VOCALISE using x-vectors
CN107481717A (en) A kind of acoustic training model method and system
CN111899757B (en) Single-channel voice separation method and system for target speaker extraction
CN106919662A (en) A kind of music recognition methods and system
CN110767210A (en) Method and device for generating personalized voice
CN110347426B (en) Intelligent release APP platform system and method thereof
CN111339302A (en) Method and device for training element classification model
CN113450761B (en) Parallel voice synthesis method and device based on variation self-encoder
CN112633175A (en) Single note real-time recognition algorithm based on multi-scale convolution neural network under complex environment
CN115602165B (en) Digital employee intelligent system based on financial system
CN111128211A (en) Voice separation method and device
CN111653270A (en) Voice processing method and device, computer readable storage medium and electronic equipment
US7805308B2 (en) Hidden trajectory modeling with differential cepstra for speech recognition
US8942978B2 (en) Parameter learning in a hidden trajectory model
CN113129908A (en) End-to-end macaque voiceprint verification method and system based on cycle frame level feature fusion
CN115985320A (en) Intelligent device control method and device, electronic device and storage medium
CN110600012B (en) Fuzzy speech semantic recognition method and system for artificial intelligence learning
US20070143104A1 (en) Learning statistically characterized resonance targets in a hidden trajectory model
CN114579724A (en) Seamless connection method and system for virtual human under various scenes

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant