Method, device, equipment and storage medium for pushing speech based on voice emotion

Abstract

The present invention discloses a speech push method, device, equipment and storage medium based on voice emotion, including: detecting the voice of the user's reply to obtain the target voice; recognizing the target voice to obtain the target emotion contained in the target voice; obtaining the target value of the target emotion based on the algorithm model; comparing the target value with the threshold to determine the target speech to be pushed. The technical solution of the present invention, after judging that the first dissuasion speech fails, will collect the voice of the user's reply, and perform emotion recognition processing on it, and then classify and refine the negative emotions, and through the corpus retrieval algorithm, corresponding weights are assigned to each negative emotion, and then the algorithm is used to calculate whether the threshold of the corresponding dissuasion speech is reached, and if it is reached, the corresponding dissuasion speech is recommended, realizing the combination of the anti-fraud system and the recommendation algorithm, and correspondingly recommending the appropriate dissuasion speech for the emotion after the user's voice is refined.

Description

Technical Field

[0001] The present invention belongs to the field of computer technology, and in particular relates to a method, device, equipment and storage medium for pushing speech based on voice emotions. Background Art

[0002] In the anti-fraud system, the existing solution is to use voice emotion recognition technology to roughly classify user emotions into three categories: negative, neutral, and positive. When the system identifies a possible fraud call, it will connect the call and push a unified dissuasion script. At this time, the user's response speech will be emotionally recognized. When the user's voice emotion is identified as negative, a fraud warning will be directly pushed to the user's relatives to prevent them from being deceived. Specifically, after the system identifies a possible fraud call and pushes a unified dissuasion script, if the user's voice emotion recognition result is negative, it will be judged that the dissuasion has failed, and a warning of possible fraud will be pushed to the user's relatives.

[0003] However, instead of giving a second dissuasion script based on the user's emotional reaction after the dissuasion script, a fraud warning is directly pushed to the user's relatives. This fails to push dissuasion scripts that are suitable for the user's current mood in a targeted manner, and thus fails to achieve a better dissuasion effect. Moreover, if a warning is pushed based on a judgment that the dissuasion has failed once, the possibility of making a wrong judgment will also increase. Summary of the Invention

[0004] The present invention aims to solve at least one of the technical problems in the related art to a certain extent. To this end, one object of the present invention is to provide a method, device, equipment and storage medium for pushing speech based on voice emotions.

[0005] In order to solve the above technical problems, the embodiments of the present invention provide the following technical solutions:

[0006] A speech push method based on voice emotion, comprising:

[0007] Detect the user's reply voice and obtain the target voice;

[0008] Recognizing the target speech to obtain the target emotion contained in the target speech;

[0009] Based on the algorithm model, obtaining a target value of the target emotion;

[0010] The target value is compared with the threshold value to determine the target words to be pushed.

[0011] Optionally, before detecting the user's reply voice and obtaining the target voice, the following steps may be performed:

[0012] Establishing a corpus, wherein the corpus includes a plurality of training samples;

[0013] Training the plurality of training samples to obtain a weight value and a feature value for each reference emotion;

[0014] Based on the weight value and characteristic value of each reference emotion, a threshold value of each training sample is obtained.

[0015] Optionally, the recognizing the target speech and obtaining the target emotion contained in the target speech includes:

[0016] Recognizing the target speech and determining the emotion category of the target speech;

[0017] If the emotion category is negative, recognizing the target speech based on the corpus to obtain a target reference speech that matches the target speech;

[0018] Based on the target reference speech, a target emotion included in the target reference speech is acquired.

[0019] Optionally, the target speech is recognized based on a corpus to obtain a target reference speech that matches the target speech, including the following formula:

[0020]

[0021] Where x i The component of the vector of the user's reply voice, y i is the component of the vector of the target reference speech, and n is the number of components in the vector.

[0022] Optionally, obtaining a target value of the target emotion based on an algorithm model includes:

[0023] Inputting the characteristic value and weight value of each emotion of the target emotion into the algorithm model;

[0024] The algorithm model calculates the characteristic value and weight value of each target emotion to obtain the target value of the target emotion.

[0025] Optionally, the algorithm model includes the following formula:

[0026]

[0027] Where: j is a neuron, χ i is the eigenvalue; is the weight value, and netj is the target value of the output.

[0028] Optionally, comparing the target value with a threshold value to determine a target speech to be pushed includes:

[0029] Comparing the target value with the threshold value to obtain a comparison result;

[0030] If the comparison result shows that the target value is greater than the threshold, the speech matching the target reference voice is pushed to the user;

[0031] If the comparison result is that the target value is smaller than the threshold value, the speech will not be pushed to the user.

[0032] An embodiment of the present invention further provides a speech technique push device based on voice emotion, comprising:

[0033] The detection module is used to detect the user's reply voice and obtain the target voice;

[0034] A recognition module, configured to recognize the target speech and obtain the target emotion contained in the target speech;

[0035] A calculation module, configured to obtain a target value of the target emotion based on an algorithm model;

[0036] The push module is used to compare the target value with the threshold value to determine the target words to be pushed.

[0037] An embodiment of the present invention further provides an electronic device, comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor implements the above-mentioned method when executing the computer program.

[0038] An embodiment of the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium includes a stored computer program, wherein when the computer program is executed, the device where the computer-readable storage medium is located is controlled to execute the method described above.

[0039] The embodiments of the present invention have the following technical effects:

[0040] The above technical solution of the present invention, (1) after determining that the first dissuasion speech fails, will collect the user's reply voice, and perform emotion recognition processing on it, and then classify and refine the negative emotions, and assign corresponding weight values ​​to each negative emotion through the corpus retrieval algorithm, and then calculate whether the threshold of the corresponding dissuasion speech is reached through the algorithm. If it is reached, the corresponding dissuasion speech is recommended, realizing the combination of the anti-fraud system and the recommendation algorithm, and recommending the corresponding dissuasion speech according to the emotion after the user's voice is refined.

[0041] (2) By combining the push algorithm with voice emotion recognition technology, more appropriate dissuasion words are pushed to users who may be defrauded, achieving a better dissuasion effect; at the same time, the problem of being unable to provide targeted dissuasion due to the uniformity of dissuasion words is avoided; in addition, the secondary dissuasion also greatly reduces the probability of misjudgment by the voice emotion recognition algorithm.

[0042] (3) MP algorithms can also continuously optimize the weight values ​​and thresholds corresponding to each emotion, thereby making the target values ​​obtained more accurate and the thresholds obtained more valuable for reference, thereby improving the accuracy of push words and the user experience.

[0043] (4) Based on different target voices, different target reference voices are matched, thereby achieving different dissuasive speech according to the user's emotions, achieving a better dissuasive effect.

[0044] Additional aspects and advantages of the present invention will be set forth in part in the description which follows and, in part, will be obvious from the description which follows, or may be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 This is a flow chart of a method for pushing speech based on voice emotions provided by an embodiment of the present invention;

[0046] Figure 2 It is a structural diagram of a speech pushing device based on voice emotions provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0047] The following describes embodiments of the present invention in detail, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, and are not to be construed as limiting the present invention.

[0048] In order to facilitate the understanding of the embodiments by those skilled in the art, some terms are explained:

[0049] (1) MP model: An abstract and simplified model constructed according to the structure and working principle of biological neurons. It is actually a modeling of a single neuron; in which each neuron is a multi-input and single-output information processing unit.

[0050] (2) BP algorithm: Back Propagation algorithm, the learning process consists of two processes: forward propagation of the signal and back propagation of the error.

[0051] like Figure 1As shown, an embodiment of the present invention provides a method for pushing speech based on voice emotions, comprising:

[0052] Step S1: Detect the user's reply voice to obtain the target voice;

[0053] In actual application scenarios, after determining that the first dissuasion technique fails, a second dissuasion technique is used;

[0054] Among them, based on the system's preset experience value, the user's reply voice is preliminarily detected to determine whether the user is making a fraudulent call;

[0055] If it is detected that the user may be making a fraudulent call, the user's reply voice will be used as the target voice for subsequent processing; if it is detected that the user is not making a fraudulent call, the user's voice will not be processed subsequently.

[0056] Step S2: Recognize the target speech and obtain the target emotion contained in the target speech;

[0057] Specifically, the step of recognizing the target speech and obtaining the target emotion contained in the target speech includes:

[0058] The target speech is recognized to determine the emotion category of the target speech; if the emotion category is negative, the target speech is recognized based on the corpus to obtain a target reference speech that matches the target speech; based on the target reference speech, a target emotion included in the target reference speech is obtained.

[0059] In actual application scenarios, after obtaining the target voice, the target voice is recognized based on the voice emotion recognition algorithm to identify the user's emotion. If the user's emotion is negative, the corpus is searched based on the target voice to initially obtain several relatively close reference voices, and then the target reference voice that is most similar to the target voice among the several reference voices is obtained; if the user's emotion is not negative, the target voice is not processed subsequently.

[0060] Generally, the target emotions contained in the target reference speech involve several emotions, such as anger, frustration, autism, tension, hatred, and anxiety.

[0061] The target speech is recognized based on the corpus to obtain a target reference speech that matches the target speech, including the following formula:

[0062]

[0063] Where x i The component of the vector of the user's reply voice, y iis the component of the vector of the reference speech, and n is the number of components in the vector.

[0064] The above embodiment of the present invention calculates the cosine value between the target speech and the speech in the corpus by vectorized cosine distance, and then obtains the target reference speech that is most similar to the target speech.

[0065] The closer the calculated cosine value is to 1 or -1, the more similar the target reference speech is to the target speech; the closer the cosine value is to 0, the less similar the target reference speech is to the target speech.

[0066] In actual application scenarios, you can obtain the target reference speech by following the steps below:

[0067] (1) Find the target reference speech and the keywords of the target speech;

[0068] (2) Take a number of keywords from the target reference speech and the target speech, combine them into a set, and calculate the frequency of the target reference speech and the target speech for the words in this set; for example, appear once, twice, etc.

[0069] (3) Generate the target reference speech and the word frequency vector of the target speech;

[0070] (4) Calculate the cosine similarity of two vectors. The closer the value is to 1 or -1, the more similar they are.

[0071] For example: segment the target reference speech and the target speech (obtain keywords in the speech), list all keywords, calculate the keyword frequency, and then obtain the keyword frequency vector:

[0072] The vector of the target speech (x1, x2, x3, x4, x5, x6);

[0073] The vector of the first target reference speech (y1, y2, y3, y4, y5, y6);

[0074] The vector of the second target reference speech (y7,y8,y9,y 10 ,y 11 ,y 12 );

[0075] The vector of the third target reference speech (y 13 ,y 14 ,y 15 ,,y 16 ,y 17 ,y 18 );

[0076] Then, the similarity between the vector of the target speech and the vector of the target reference speech is calculated;

[0077] The vector of the above-mentioned target speech and the first target reference speech, the vector of the target speech and the second target reference speech, and the vector of the target speech and the third target reference speech are substituted into the formula respectively, and the cosine values ​​obtained are 0.89, 0.91, and 0.66 respectively, indicating that the target speech and the second reference speech are highly similar, that is, the second reference speech is the target reference speech.

[0078] Then, based on the corpus, several emotions (such as anger, irritability, anxiety, etc.) contained in the target reference speech are obtained, and the weight values ​​and feature values ​​corresponding to each of these emotions are obtained based on the corpus.

[0079] The embodiments of the present invention, by combining a push algorithm with voice emotion recognition technology, push more appropriate dissuasion words to users who may be defrauded, thereby achieving a better dissuasion effect; at the same time, it avoids the problem of inability to provide targeted dissuasion due to the uniformity of dissuasion words; in addition, secondary dissuasion also greatly reduces the probability of misjudgment of the voice emotion recognition algorithm.

[0080] Step S3: obtaining a target value of the target emotion based on the algorithm model;

[0081] Specifically, the method for obtaining the threshold and weight value includes: establishing a corpus, the corpus including multiple training samples; training the multiple training samples to obtain the weight value and characteristic value of each reference emotion; and obtaining the threshold of each training sample based on the weight value and characteristic value of each reference emotion.

[0082] Among them, the algorithm model of the embodiment of the present invention has been established before detecting the voice of the user's reply, so as to save the time cost of obtaining the target value of the target emotion. In addition, the corpus can be updated in real time as needed to improve applicability.

[0083] In actual application scenarios, the MP algorithm commonly used in neural networks (or the MP model of neural networks) is used to learn and train multiple training samples.

[0084] For example, set the number of training times to 100 times (can be set according to actual conditions), and train multiple training samples (a segment of speech with negative emotions), obtain the emotions contained in each training sample, and record the weight value corresponding to each emotion and the threshold value corresponding to each training sample, and mark the weight value on the corresponding emotion in the corpus for subsequent retrieval and use.

[0085] In an embodiment of the present invention, the above-mentioned MP algorithm can also continuously optimize the weight value and threshold value corresponding to each emotion, thereby making the obtained target value more accurate and the obtained threshold value more valuable for reference, thereby improving the accuracy of the push speech and improving the user experience.

[0086] Wherein, obtaining the target value of the target emotion based on the algorithm model includes:

[0087] Inputting the characteristic value and weight value of each emotion of the target emotion into the algorithm model;

[0088] The algorithm model calculates the characteristic value and weight value of each target emotion to obtain the target value of the target emotion.

[0089] Optionally, the algorithm model includes the following formula:

[0090]

[0091] Where: j is a neuron, χ i is the eigenvalue; is the weight value, and netj is the target value of the output.

[0092] Specifically, in the algorithm model, the weight value accumulates and integrates all the eigenvalues ​​based on their corresponding weight values ​​to output the target value.

[0093] For example, in the corpus, according to the results of MP algorithm training, the threshold of the dissuasive speech "Please don't be anxious, let's talk slowly and calmly." is 2. The target speech is calculated by the above formula to obtain the target emotion, which includes anger, irritability, and anxiety. The weight values ​​corresponding to each emotion are 0.2, 0.2, and 0.6. The weight values ​​corresponding to each emotion included in the target emotion and the signal of the target speech are substituted into the above formula, and the target value of the target speech is 2.1. It can be seen that the target value of the target speech is greater than the threshold, so the dissuasive speech "Please don't be anxious, let's talk slowly and calmly" will be pushed to the user.

[0094] In the corpus, according to the results of MP algorithm training, the threshold of the dissuasive speech "Don't keep it in your heart, tell me what you want to say, and I will listen patiently" is 3. If the target speech is calculated by the above formula to obtain the target emotion, the target emotions include autism and depression, where the weight values ​​corresponding to each emotion are 0.6 and 0.4 respectively. Substituting the weight values ​​and characteristic values ​​corresponding to each emotion included in the target emotion into the above formula, the target value of the target speech is 3.2. It can be seen that the target value of the target speech is greater than the threshold. The dissuasive speech finally pushed by the above algorithm may be: "Don't keep it in your heart, tell me what you want to say, and I will listen patiently." This type of speech.

[0095] The embodiments of the present invention match different target reference voices based on different target voices, thereby achieving the goal of pushing different dissuasive speech techniques according to the user's emotions, achieving a better dissuasive effect.

[0096] In an optional embodiment of the present invention, the target value can be obtained by processing the feature values ​​and weight values ​​corresponding to each emotion included in the target speech based on the BP algorithm. The embodiment of the present invention does not specifically limit the specific execution process of the BP algorithm.

[0097] Step S4: Compare the target value with the threshold value to determine the target speech to be pushed.

[0098] Specifically, comparing the target value with a threshold value to determine the target speech to be pushed includes:

[0099] Comparing the target value with the threshold value to obtain a comparison result;

[0100] If the comparison result shows that the target value is greater than the threshold, the speech matching the target reference voice is pushed to the user;

[0101] If the comparison result is that the target value is smaller than the threshold value, the speech will not be pushed to the user.

[0102] In the corpus, each reference voice corresponds to a threshold, and each reference voice corresponds to a dissuasive speech. If the target value of the target voice obtained by calculation is greater than the threshold of the target reference voice, the dissuasive speech matching the target reference voice will be pushed to the user; if the target value of the target voice obtained by calculation is not greater than the threshold of the target reference voice, the dissuasive speech matching the target reference voice will not be pushed to the user.

[0103] In addition, the MP algorithm can continuously train the corpus or continuously obtain new training samples based on the settings, and train multiple training samples to achieve more accurate optimization of the weight value and threshold corresponding to each emotion in the corpus.

[0104] In an embodiment of the present invention, after determining that the first dissuasion speech fails, the user's reply voice will be collected and processed for emotion recognition, and the negative emotions will be classified and refined. A corpus retrieval algorithm will be used to assign a corresponding weight value to each negative emotion. The algorithm will then be used to calculate whether the threshold of the corresponding dissuasion speech is reached. If so, the corresponding dissuasion speech will be recommended. This realizes the combination of the anti-fraud system and the recommendation algorithm, and recommends corresponding dissuasion speech according to the emotion after the user's voice is refined.

[0105] like Figure 2 As shown, an embodiment of the present invention further provides a speech technique push device 200 based on speech emotion, comprising:

[0106] Detection module 201, used to detect the user's reply voice and obtain the target voice;

[0107] Recognition module 202, configured to recognize the target speech and obtain the target emotion contained in the target speech;

[0108] A calculation module 203 is used to obtain a target value of the target emotion based on an algorithm model;

[0109] The push module 204 is used to compare the target value with a threshold value to determine the target speech to be pushed.

[0110] In an optional embodiment of the present invention, the above-mentioned apparatus further includes a corpus retrieval algorithm for calculating the similarity between the target speech and the reference speech, and further determining the target reference speech.

[0111] In an optional embodiment of the present invention, the above-mentioned device further includes a target value calculation module, which is used to calculate the sum of the product of the characteristic value and the weight value of the target emotion and output the target value.

[0112] Optionally, the method for obtaining the threshold and weight value includes:

[0113] Establishing a corpus, wherein the corpus includes a plurality of training samples;

[0114] Training the plurality of training samples to obtain a weight value and a feature value for each reference emotion;

[0115] Based on the weight value and the characteristic value of each reference emotion, a threshold value of each training sample is obtained.

[0116] Optionally, the recognizing the target speech and obtaining the target emotion contained in the target speech includes:

[0117] Recognizing the target speech and determining the emotion category of the target speech;

[0118] If the emotion category is negative, the target speech is recognized based on the corpus to obtain a target reference speech that matches the target speech; and based on the target reference speech, a target emotion included in the target reference speech is obtained.

[0119] Optionally, the target speech is recognized based on a corpus to obtain a reference speech that matches the target speech, including the following formula:

[0120]

[0121] Where x i The component of the vector of the user's reply voice, y i is the component of the vector of the reference speech, and n is the number of components in the vector.

[0122] Optionally, obtaining a target value of the target emotion based on an algorithm model includes:

[0123] Inputting the characteristic value and weight value of each emotion of the target emotion into the algorithm model;

[0124] The algorithm model calculates the characteristic value and weight value of each target emotion to obtain the target value of the target emotion.

[0125] Optionally, the algorithm model includes the following formula:

[0126]

[0127] Where: j is a neuron, χ i is the eigenvalue; is the weight value, and netj is the target value of the output.

[0128] Optionally, comparing the target value with a threshold value to determine a target speech to be pushed includes:

[0129] Comparing the target value with the threshold value to obtain a comparison result;

[0130] If the comparison result shows that the target value is greater than the threshold, the speech matching the reference speech is pushed to the user;

[0131] If the comparison result is that the target value is smaller than the threshold value, the speech will not be pushed to the user.

[0132] An embodiment of the present invention further provides an electronic device, comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor implements the above-mentioned method when executing the computer program.

[0133] An embodiment of the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium includes a stored computer program, wherein when the computer program is executed, the device where the computer-readable storage medium is located is controlled to execute the method described above.

[0134] In addition, other structures and functions of the device in the embodiment of the present invention are known to those skilled in the art and are not described here in detail to reduce redundancy.

[0135] It should be noted that the logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing the logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device). For purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (non-exhaustive list) of computer-readable media include the following: an electrical connection with one or more wires (electronic device), a portable computer disk cartridge (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and programmable read-only memory (EPROM or flash memory), fiber optic devices, and portable compact disc read-only memory (CDROM). Furthermore, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting or processing it in another suitable manner if necessary, and then storing it in a computer memory.

[0136] It should be understood that various parts of the present invention can be implemented using hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, an application-specific integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0137] Throughout this specification, reference to terms such as "one embodiment," "some embodiments," "examples," "specific examples," or "some examples" means that a specific feature, structure, material, or characteristic described in conjunction with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, schematic representations of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

[0138] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", "clockwise", "counterclockwise", "axial", "radial", "circumferential" and the like to indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as limiting the present invention.

[0139] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of the technical features being referred to. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one such feature. In the description of the present invention, "plurality" means at least two, such as two, three, etc., unless otherwise specifically defined.

[0140] In the present invention, unless otherwise specified or limited, the terms "installed," "connected," "connect," "fixed," etc. should be understood in a broad sense. For example, they can refer to fixed connection, detachable connection, or integration; mechanical connection, electrical connection; direct connection, or indirect connection through an intermediate medium; internal communication between two components, or interaction between two components, unless otherwise specified. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on specific circumstances.

[0141] In the present invention, unless otherwise expressly specified or limited, when a first feature is "above" or "below" a second feature, it may mean that the first and second features are in direct contact, or that the first and second features are in indirect contact through an intermediary. Furthermore, when a first feature is "above," "above," or "above" a second feature, it may mean that the first feature is directly above or diagonally above the second feature, or simply means that the first feature is at a higher level than the second feature. When a first feature is "below," "below," or "below" a second feature, it may mean that the first feature is directly below or diagonally below the second feature, or simply means that the first feature is at a lower level than the second feature.

[0142] Although the embodiments of the present invention have been shown and described above, it will be understood that the above embodiments are illustrative and are not to be construed as limitations on the present invention. A person skilled in the art may change, modify, replace and modify the above embodiments within the scope of the present invention.

Claims

1. A speech technique push method based on voice emotion, characterized in that: include: Detect the user's reply voice and obtain the target voice; Recognizing the target speech and obtaining the target emotion contained in the target speech; Based on the algorithm model, obtaining a target value of the target emotion; Compare the target value with the threshold value to determine the target words to be pushed; Before detecting the user's reply voice and obtaining the target voice, the following steps are also included: Establishing a corpus, wherein the corpus includes a plurality of training samples; Training the plurality of training samples to obtain a weight value and a feature value of each reference emotion; Based on the weight value and characteristic value of each reference emotion, a threshold value of each training sample is obtained; The step of recognizing the target speech and obtaining the target emotion contained in the target speech includes: Recognizing the target speech and determining the emotion category of the target speech; If the emotion category is negative, recognizing the target speech based on the corpus to obtain a target reference speech that matches the target speech; Based on the target reference speech, a target emotion included in the target reference speech is acquired.

2. The method according to claim 1, characterized in that Recognizing the target speech based on the corpus and acquiring a target reference speech matching the target speech includes the following formula: In the formula, x i The component of the vector of the user's reply voice, y i is the component of the vector of the target reference speech, and n is the number of components in the vector.

3. The method according to claim 1, characterized in that The step of obtaining a target value of the target emotion based on the algorithm model includes: Inputting the characteristic value and weight value of each emotion of the target emotion into the algorithm model; The algorithm model calculates the characteristic value and weight value of each emotion of the target emotion to obtain the target value of the target emotion.

4. The method according to claim 1, characterized in that: The algorithm model includes the following formula: Where: j is a neuron, χ i is the characteristic value; is the weight value, and netj is the target value of the output.

5. The method according to claim 1, characterized in that The step of comparing the target value with the threshold value to determine the target speech to be pushed includes: Comparing the target value with the threshold value to obtain a comparison result; If the comparison result is that the target value is greater than the threshold, the speech matching the target reference voice is pushed to the user; If the comparison result is that the target value is smaller than the threshold value, the speech will not be pushed to the user.

6. A speech technique push device based on voice emotion, characterized in that: include: The detection module is used to detect the voice of the user's reply and obtain the target voice; A recognition module, used to recognize the target speech and obtain the target emotion contained in the target speech; A calculation module, used for obtaining a target value of the target emotion based on an algorithm model; A push module, used to compare the target value with a threshold value to determine a target phrase to be pushed; Before detecting the voice replied by the user and obtaining the target voice, the device is also used to: Establishing a corpus, wherein the corpus includes a plurality of training samples; Training the plurality of training samples to obtain a weight value and a feature value of each reference emotion; Based on the weight value and characteristic value of each reference emotion, a threshold value of each training sample is obtained; The step of recognizing the target speech and obtaining the target emotion contained in the target speech includes: Recognizing the target speech and determining the emotion category of the target speech; If the emotion category is negative, recognizing the target speech based on the corpus to obtain a target reference speech that matches the target speech; Based on the target reference speech, a target emotion included in the target reference speech is acquired.

7. An electronic device, characterized in that: The method comprises a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor implements the method according to any one of claims 1 to 5 when executing the computer program.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored computer program, wherein when the computer program is executed, the device where the computer-readable storage medium is located is controlled to execute the method according to any one of claims 1 to 5.

Images