Bluetooth sound box control method and system based on AI voice analysis

Abstract

The invention discloses a Bluetooth sound box control method and system based on AI voice analysis, relates to the technical field of Bluetooth sound box control, and solves the technical problem that the playing volume is difficult to adapt to user requirements by manually adjusting the volume of a Bluetooth sound box in the prior art. When the Bluetooth sound box is controlled, a user firstly generates a wake-up audio through a wake-up word, and the Bluetooth sound box recognizes a corresponding dialect according to the wake-up audio and calls an audio recognition library corresponding to the dialect; the audio recognition library is used for recognizing the subsequently received control audio, determining the feedback volume according to the audio attribute of the control audio, generating a control instruction according to the audio content of the control audio, and finally sending feedback information according to the feedback volume; the feedback volume is automatically adjusted according to the control audio received in real time, so that the volume emitted by the Bluetooth sound box can be effectively prevented from being too large or too small, the interaction quality between the Bluetooth sound box and a user is improved, and the user experience is improved.

Description

Technical Field

[0001] The present application relates to the field of Bluetooth speaker control and involves AI speech recognition and analysis technology, specifically a Bluetooth speaker control method and system based on AI speech analysis. Background Art

[0002] A Bluetooth speaker is an audio device that uses a built-in Bluetooth chip to replace the traditional wire connection with a Bluetooth connection. Bluetooth speakers can achieve audio playback, question-and-answer interaction, and local device control through control commands or voice interaction. They are gradually being valued by consumers for their compact appearance and rich functions.

[0003] When used indoors, Bluetooth speakers are usually placed in a fixed position. Their main function is to interact with users and help users control smart home devices. Whether it is interaction or control, Bluetooth speakers should provide users with timely feedback. However, the volume adjustment of Bluetooth speakers is currently mainly manual control and smartphone control, and because the position of Bluetooth speakers is relatively fixed, it is difficult for the playback volume to adapt to user needs, which will seriously affect the interactive effect of Bluetooth speakers.

[0004] This application provides a Bluetooth speaker control method and system based on AI voice analysis to solve the above technical problems. Summary of the invention

[0005] The present application aims to solve at least one of the technical problems existing in the prior art; to this end, the present application proposes a Bluetooth speaker control method and system based on AI voice analysis, which is used to solve the technical problem that the volume of the Bluetooth speaker is difficult to adapt to user needs by manually adjusting the volume of the prior art.

[0006] To achieve the above objectives, the first aspect of the present application provides a Bluetooth speaker control method based on AI voice analysis, comprising: Receive wake-up audio, recognize wake-up audio and call audio recognition library; receiving control audio, identifying audio attributes of the control audio, and automatically adjusting the feedback volume of the Bluetooth speaker according to the audio attributes; wherein the audio attributes include direction and distance; and, The control audio is recognized through the audio recognition library, and a control instruction is generated according to the recognition result; the operation is completed according to the control instruction, and feedback information is issued according to the feedback volume.

[0007] Preferably, before receiving the wake-up audio, setting the wake-up word of the Bluetooth speaker includes: Determine a basic word and several candidate words; combine the basic word and several candidate words in sequence to form a candidate phrase; wherein the basic word is a required word set by the Bluetooth speaker manufacturer; Perform audio tests on several candidate phrases, and select at least one candidate phrase as a wake-up word based on the test results; wherein the audio test includes a wake-up rate test, a stability test, and a discrimination rate test.

[0008] Preferably, audio tests are performed on several candidate phrases, including: Test the arousal rate and stability of several alternative phrases; The test obtains audio signals of different dialects corresponding to each candidate phrase, and calculates the discrimination rate of the corresponding candidate phrase according to the similarity between the audio signals corresponding to different dialects; Based on the wake-up rate, stability and discrimination rate, the alternative phrase with the best comprehensive performance is selected as the wake-up word.

[0009] Preferably, identifying audio properties of the control audio includes: Calculate the time difference between the control audio and the wake-up audio; When the time difference is greater than the set time threshold, a prompt is given to wake up the device again; otherwise, the relative position of the control audio and the Bluetooth speaker is identified; Gets the audio properties that control the audio based on the relative position.

[0010] Preferably, automatically adjusting the feedback volume of the Bluetooth speaker according to the audio attribute includes: Retrieving a property-volume curve; wherein the property-volume curve represents a corresponding relationship between audio properties and an optimal volume, and is constructed through simulation testing; The audio attribute is input into the attribute-volume curve, the optimal volume corresponding to the audio attribute is calculated, and the optimal volume is used as the feedback volume.

[0011] Preferably, before calculating the time difference between the control audio and the wake-up audio, determining the dialect similarity between the two includes: Determine the dialect similarity between the control audio and the wake-up audio; When the dialect similarity is greater than the similarity threshold, the control audio is recognized by the audio recognition library; otherwise, the control audio is analyzed to re-match the audio recognition library.

[0012] Preferably, when several control audios are received, the timbre characteristics of each control audio are identified; the timbre characteristics are compared with the timbre characteristics of a preset user, the control audio with consistent comparison is used as the target audio, and the target audio is analyzed to generate a control instruction; wherein the preset user has priority.

[0013] Preferably, the feedback volume is determined according to a number of audio properties of the control audio, including: Identify several audio attributes of the control audio, select the audio attribute closest to the Bluetooth speaker from the several audio attributes as the first target attribute, and use the audio attribute of the target audio as the second target attribute; Volume one is determined according to target attribute one and the attribute-volume curve, and volume two is determined according to target attribute two and the attribute-volume curve; feedback volume is determined based on volume one and volume two; volume one is the reminder volume.

[0014] Preferably, determining the feedback volume based on the volume one and the volume two includes: Set volume 1 as the reminder volume and volume 2 as the target volume; After setting the transition between the alert volume and the target volume, the feedback volume is generated.

[0015] The second aspect of the present application provides a Bluetooth speaker control system based on AI voice analysis, which is applied to a Bluetooth speaker, including: Data acquisition module: used to receive wake-up audio and control audio; Speech analysis module: used to analyze the wake-up audio and call the audio recognition library, as well as determine the feedback volume according to the audio properties of the control audio and send feedback information according to the feedback volume.

[0016] Compared with the prior art, the beneficial effects of this application are: 1. In this application, when controlling a Bluetooth speaker, the user first generates a wake-up audio through a wake-up word. The Bluetooth speaker recognizes the corresponding dialect based on the wake-up audio and calls the audio recognition library corresponding to the dialect. The control audio subsequently received is recognized through the audio recognition library, and the feedback volume is determined based on the audio properties of the control audio. A control instruction is generated based on the audio content of the control audio, and finally feedback information is sent out at the feedback volume. The feedback volume of this application is automatically adjusted according to the control audio received in real time, which can effectively avoid the volume emitted by the Bluetooth speaker being too loud or too low, improve the interaction quality between the Bluetooth speaker and the user, and enhance the user experience.

[0017] 2. This application obtains several alternative phrases based on the combination of basic words and alternative words, conducts audio tests on several alternative phrases in different dialects, calculates the comprehensive score of each alternative phrase based on the wake-up rate, stability and discrimination rate, and selects one or more with the highest comprehensive score as the wake-up word. Generating multiple alternative phrases based on basic words and alternative words can meet the user's customization needs and the manufacturer's advertising needs; testing the wake-up rate, stability and discrimination rate of alternative phrases in different dialects, selecting the best performing alternative phrase as the wake-up word, calling the audio recognition library based on the wake-up word to identify subsequent control audio, can improve the voice interaction accuracy of Bluetooth speakers in different dialect areas.

[0018] 3. After verifying the interval time and dialect consistency between the control audio and the wake-up audio, this application identifies the audio attributes of the control audio, calls the attribute-volume curve according to the scene in which the Bluetooth speaker is located, obtains the optimal volume corresponding to the audio attribute, and uses the optimal volume as the feedback volume. This application determines the optimal volume by controlling the audio, which is equivalent to determining the feedback volume according to the user's latest location. It can provide the most appropriate feedback volume for mobile users, and when the feedback volume is obtained using the two parameters of direction and distance, the scene applicability can be improved, ensuring that the feedback volume can meet user requirements without the need for users to adjust frequently. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0020] Figure 1 This is a schematic diagram of the method flow of the Bluetooth speaker control method in Embodiment 1 of the present application; Figure 2 This is a schematic diagram of the determination process for identifying wake-up words in various dialects in Embodiment 2 of the present application; Figure 3 This is a flow chart of determining the feedback volume according to the audio properties of the controlled audio in the third embodiment of the present application; Figure 4 This is a flow chart of determining the feedback volume when the Bluetooth speaker receives a plurality of control audios in the fourth embodiment of the present application; Figure 5 This is a schematic diagram of the system principle of the Bluetooth speaker control system in this application. DETAILED DESCRIPTION

[0021] The technical solution of the present application will be described clearly and completely in conjunction with the embodiments below. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present application.

[0022] Bluetooth speakers are increasingly used in people's daily lives, such as being placed indoors to play music, to have fun communicating with users, and even to use Bluetooth speakers as control centers to control smart home devices. Regardless of playing music, communicating, or controlling devices, Bluetooth speakers have corresponding outputs, but when Bluetooth speakers are used indoors, their positions are relatively fixed, and their volume control methods are also relatively fixed. They are either manually controlled through buttons on the Bluetooth speakers, or controlled by smart terminals (such as smartphones, tablets, etc.) connected to the Bluetooth speakers. Once the volume of the Bluetooth speakers is controlled, it is determined. Unless the volume is readjusted, the volume of the Bluetooth speakers remains unchanged and will not be adjusted automatically. This results in the output volume of the Bluetooth speakers remaining unchanged no matter how far away they are from the user. Obviously, when the distance between the two is far, it is difficult for users to receive feedback from the Bluetooth speakers. When the distance between the two is close, users may be frightened by the feedback from the Bluetooth speakers, thereby affecting the interaction effect and reducing the user's interaction experience.

[0023] Embodiment 1 The Bluetooth speaker control method based on AI voice analysis provided in this embodiment controls the feedback volume by identifying the direction and distance of the audio signal, thereby ensuring that the feedback volume can clearly convey the feedback information to the user while avoiding the feedback volume being too loud and affecting the user experience.

[0024] See also Figure 1 The first embodiment of the present application provides a Bluetooth speaker control method based on AI voice analysis, including: Receive wake-up audio, recognize wake-up audio and call audio recognition library; receiving control audio, identifying audio attributes of the control audio; automatically adjusting the feedback volume of the Bluetooth speaker according to the audio attributes; wherein the audio attributes include direction and distance; and, The control audio is recognized through the audio recognition library, and a control instruction is generated according to the recognition result; the operation is completed according to the control instruction, and feedback information is issued according to the feedback volume.

[0025] The Bluetooth speaker control method provided in this embodiment involves two audio signals, one is the wake-up audio and the other is the control audio. The wake-up audio is used to remind the Bluetooth speaker to accurately receive the subsequent control audio, while the control audio contains the control content required by the user. Both the wake-up audio and the control audio come from the user. It should be noted that the wake-up audio generally precedes the control audio, and the audio coming from the user can mean directly from the user, or indirectly from the user.

[0026] During daily use, the Bluetooth speaker may be in various types of dialect areas, that is, it is necessary to deal with various types of dialect scenarios. In order to ensure that the dialect can be accurately identified so as to accurately identify the subsequent control audio, this embodiment determines the corresponding dialect type through a specific wake-up audio, calls the corresponding audio recognition library according to the dialect type, and identifies the subsequent control audio through the audio recognition library. There are multiple audio recognition libraries built into the Bluetooth speaker or in the cloud storage connected to it, and each audio recognition library is used to identify a dialect. After determining the type of dialect used by the user, calling the corresponding audio recognition library for identifying the control audio can improve the recognition accuracy of the control audio, and calling the audio recognition library in advance can improve the recognition efficiency of the control audio, avoiding the situation where the control audio has been received but the audio recognition library call has not been completed. It should be noted that the audio recognition library is pre-trained with various dialect training data sets, which can be an existing commercial model or an artificial intelligence model trained by self-collected data. It should be noted that the speech recognition of this application is based on existing speech recognition methods and AI large model technology.

[0027] After receiving the control audio, the audio properties of the control audio are first identified, which specifically include the relative distance and relative direction between the control audio emission point and the Bluetooth speaker. The relative direction can use Bluetooth DirectionFinding technology to identify the direction of the user's voice. This technology includes two methods, Angle of Arrival (AoA) and Angle of Departure (AoD), which determine the direction of the signal by measuring the phase difference of the signal. For example, in the AoA method, the receiving device (speaker) uses an antenna array to measure the phase difference of a special signal sent by a single transmitting antenna to estimate the relative direction of the signal. In addition, Bluetooth® Channel Sounding technology uses phase-based ranging (PBR) to achieve high-precision distance measurement (HADM).

[0028] In some other preferred embodiments, the relative distance may be measured by a built-in distance measuring sensor based on the relative direction, such as by using an infrared sensor to identify the relative distance to the user along the relative direction.

[0029] The audio attributes identified by the above method can determine the relative position between the user and the Bluetooth speaker (mainly including relative direction and relative distance), and the feedback volume after the Bluetooth speaker is set according to the relative position. Each time the control audio is received, the feedback volume is adjusted according to the audio attributes of the control audio (when the audio attributes have not changed or have changed slightly, frequent adjustment is not required), so that the volume of the Bluetooth speaker can be automatically set. This can solve the problem that the volume of the Bluetooth speaker is fixedly set through the existing solution, and the user may not be able to hear the feedback information clearly when the relative distance is far, and the volume may be too loud when the relative distance is close, affecting the user experience.

[0030] While setting the feedback volume according to the audio properties, the audio recognition library is used to identify the content of the control audio, and a control instruction is generated based on the identified content, and the relevant operation is performed according to the control instruction. If the operation is music playback, smart question and answer, etc., which requires audio playback, the aforementioned feedback volume is the playback volume; if the operation is to control other devices, feedback information is issued according to the feedback volume after the control is completed. Of course, if the user sets that no feedback information is required, no feedback information will be issued.

[0031] In this embodiment, when controlling the Bluetooth speaker, the user first generates a wake-up audio through a wake-up word, and the Bluetooth speaker recognizes the corresponding dialect based on the wake-up audio, and calls the audio recognition library corresponding to the dialect; the control audio subsequently received is recognized through the audio recognition library, the feedback volume is determined according to the audio properties of the control audio, the control instruction is generated according to the audio content of the control audio, and finally the feedback information is sent out with the feedback volume. The feedback volume of this embodiment is automatically adjusted according to the control audio received in real time, which can effectively avoid the volume emitted by the Bluetooth speaker being too loud or too low, improve the interaction quality between the Bluetooth speaker and the user, and enhance the user experience.

[0032] Embodiment 2 Based on the first embodiment, this embodiment provides a method for determining a wake-up word to improve the recognizability of the wake-up audio, improve the matching degree of the audio recognition library, and thus improve the recognition accuracy of the control audio. Figure 2 , as follows: Before receiving the wake-up audio, set the wake-up word of the Bluetooth speaker, including: Determine a basic word and several candidate words; combine the basic word with several candidate words in sequence to form candidate phrases; perform audio tests on the several candidate phrases, and select at least one candidate phrase as a wake-up word based on the test results.

[0033] The wake-up word can be selected from a candidate phrase consisting of several candidate words and basic words input by the user after the Bluetooth speaker is activated, or at least one candidate phrase that meets the requirements is determined in advance when the Bluetooth speaker leaves the factory, and the user selects one from the built-in candidate phrases as the wake-up word. In other words, the wake-up word test process can be performed after the Bluetooth speaker is activated, or it can be completed before the Bluetooth speaker leaves the factory.

[0034] The above basic words are fixed words, generally the code name or nickname of the Bluetooth speaker. The alternative words are phrases set by the manufacturer or user. The alternative words and basic words should have coherent semantics after forming a phrase, and preferably have positive meaning. There is no regulation on the order of alternative words and basic words in the alternative phrase, that is, one alternative word and one basic word may generate two alternative phrases by adjusting the order.

[0035] After the basic words and candidate words form several candidate phrases, audio tests need to be performed on each candidate phrase to ensure that the candidate phrases with the required wake-up rate and stability and that can effectively distinguish dialects are selected as wake-up words. Audio tests on several candidate phrases include the following steps: Test the wake-up rate and stability of several candidate phrases; obtain the audio signals of different dialects corresponding to each candidate phrase, and calculate the discrimination rate of the corresponding candidate phrase according to the similarity between the corresponding audio signals of different dialects; select the candidate phrase with the best comprehensive performance as the wake-up word based on the wake-up rate, stability and discrimination rate.

[0036] The wake-up rate refers to the probability that the candidate phrase is correctly recognized, that is, the proportion of the Bluetooth speaker that is successfully awakened among all the pronunciations that are actually candidate phrases. Stability mainly refers to the recognition stability in the awakening and non-awakening states, which can evaluate the Bluetooth speaker's ability to recognize candidate phrases under long-term operation. In addition to the awakening rate and stability improved in this embodiment, it can also include false awakening rate, response time, false rejection rate, etc.

[0037] The above discrimination rate is used to indicate the ability of several candidate phrases to distinguish various dialects, which can be calculated by the similarity between audio signals. Please refer to the following steps: Extracting a number of audio signals in different dialects for each candidate phrase, and extracting MFCC (Mel-Frequency Cepstral Coefficients) coefficients of the audio signals; the calculation method of the MFCC coefficients is widely documented in existing schemes, and the calculation process thereof will not be described in detail here; Calculate the similarity of any two MFCC coefficients. The higher the similarity, the less obvious the distinction between the two dialects is for the corresponding candidate phrase (which can be distinguished by the set similarity threshold). The proportion of dialects that are clearly distinguished in the same candidate phrase can be statistically calculated as the distinction rate. The higher the distinction rate, the better the ability of the corresponding candidate phrase to distinguish different dialects is.

[0038] It should be noted that the above similarity calculation methods include Euclidean distance and cosine similarity. Euclidean distance: calculates the Euclidean distance between two sets of MFCC coefficients. The smaller the distance, the higher the similarity. Cosine similarity: calculates the cosine similarity between two sets of MFCC coefficients. The closer the cosine value is to 1, the higher the similarity.

[0039] In some other preferred embodiments, Mel-spectrum may be used to compare audio similarity, such as extracting the Mel-spectrum of the audio signal first, and calculating the sum of square differences between the Mel-spectrums of two audio signals to evaluate the similarity between the two.

[0040] Finally, you need to select the candidate phrase with the best overall performance as the wake-up word based on the wake-up rate, stability, and discrimination rate. You can refer to the following steps: 1. Make all evaluation indicators positive, that is, ensure that all indicators are as high as possible, and then standardize the data to eliminate the influence of different indicator dimensions. The standardization formula is: ;in, Indicates The alternative phrases are in The original data on the evaluation indicators, is the standardized data.

[0041] 2. Calculate positive and negative ideal solutions: For each evaluation indicator, calculate the maximum and minimum values ​​of all candidate phrases and use them as positive ideal solutions. and negative ideal solution .

[0042] 3. Calculate the Euclidean distance between the candidate phrase and the positive ideal solution and the negative ideal solution respectively: in, It is The distance between the candidate phrases and the ideal solution, It is The distance between the candidate phrases and the negative ideal solution.

[0043] 4. Calculate the comprehensive score of each candidate phrase , ; Comprehensive score The higher it is, the closer the corresponding alternative phrase is to the ideal solution, and the better the effect is.

[0044] It can be understood that the weight of each evaluation indicator in the above steps needs to be assigned according to its importance. The weight can be determined by expert scoring method, entropy method, hierarchical analysis method and other methods. The weighted standardized evaluation indicator data is used to calculate the comprehensive score.

[0045] After calculating the comprehensive scores of the candidate phrases, the phrase with the highest comprehensive score can be selected as the wake-up word. Considering the different preferences of users, several candidate phrases with the highest comprehensive scores can be selected, and the user can select one of them as the wake-up word.

[0046] This embodiment obtains several candidate phrases based on the combination of basic words and candidate words, performs audio tests on several candidate phrases in different dialects, calculates the comprehensive score of each candidate phrase based on the wake-up rate, stability and discrimination rate, and selects one or more with the highest comprehensive score as the wake-up word. Generating multiple candidate phrases based on basic words and candidate words can meet the user's customization needs and the manufacturer's advertising needs; testing the wake-up rate, stability and discrimination rate of the candidate phrases in different dialects, selecting the best performing candidate phrase as the wake-up word, calling the audio recognition library based on the wake-up word to identify the subsequent control audio, can improve the voice interaction accuracy of the Bluetooth speaker in different dialect areas.

[0047] Embodiment 3 Based on the first or second embodiment, this embodiment provides a method for determining the feedback volume. This method not only considers the relative position between the user who sends the wake-up audio and the Bluetooth speaker, but also considers the regional characteristics of the area where the Bluetooth speaker is placed. After comprehensive analysis, the optimal feedback volume is guaranteed to avoid the influence of environmental noise on the user's reception of feedback information. Figure 3 , please refer to the following steps for details: First identify the audio properties that control the audio, including: Calculate the time difference between the control audio and the wake-up audio; when the time difference is greater than the set time threshold, prompt to re-wake up the device; otherwise, identify the relative position of the control audio and the Bluetooth speaker; and obtain the audio properties of the control audio according to the relative position.

[0048] Before determining the feedback volume, it is also necessary to first determine whether the control audio needs to be fed back, that is, whether the control audio is connected to the wake-up audio. Therefore, the time difference between the control audio and the wake-up audio is calculated to determine whether the two are connected. If the time difference is greater than the set time threshold, the user is prompted to wake up the device again. If the time difference is less than the set time threshold, the audio properties of the control audio are identified. It should be noted that when prompting the user to wake up the device again, the volume of the Bluetooth speaker can be determined according to the audio properties of the wake-up audio to ensure that the prompt volume allows the user to clearly receive the message.

[0049] In some other preferred embodiments, before calculating the time difference between the control audio and the wake-up audio, the dialect similarity between the two can also be determined, including: determining the dialect similarity between the control audio and the wake-up audio; when the dialect similarity is greater than the similarity threshold, identifying the control audio through the audio recognition library; otherwise, analyzing the control audio to re-match the audio recognition library.

[0050] After receiving the control audio, the dialect similarity between the control audio and the wake-up audio is calculated to ensure that the audio recognition library called according to the wake-up audio is suitable for the control audio. If the control audio and the wake-up audio do not belong to the same dialect, it is necessary to redetermine and call the audio recognition library according to the dialect corresponding to the control audio to improve the recognition accuracy of the control audio. In some other preferred embodiments, when the dialects of the control audio and the wake-up audio are inconsistent, a reminder can also be provided to re-wake up the Bluetooth speaker.

[0051] After identifying the audio attribute that controls the audio, retrieve the attribute-volume curve in the Bluetooth speaker, input the relevant parameters in the audio attribute into the attribute-volume curve, obtain the optimal volume of the audio attribute, and use the optimal volume as the feedback volume. It should be noted that the audio attribute in the attribute-volume curve is used as the independent variable, and its optimal volume is used as the dependent variable. The curve can be obtained by simulating a large amount of data, and the independent variable in the attribute-volume curve in general scenarios is the distance, and the feedback volume is determined according to the distance.

[0052] In some other preferred embodiments, the independent variables in the attribute-volume curve may include two parameters, direction and distance. This type of curve is suitable for scenes with more background noise or more complex spatial layout. The attribute-volume curve can self-learn the optimal volume according to the volume adjusted by daily users, and this type of attribute-variable curve can be established when there is enough data. Of course, it is also possible to simulate and train attribute-variable curves for a variety of special scenarios in the laboratory, and select according to the actual scenario of the Bluetooth speaker.

[0053] It is worth noting that the feedback volume can also be determined based on the audio properties of the wake-up audio, because the direction and distance between the user and the Bluetooth speaker can also be identified based on the wake-up audio. However, when the user actually uses the Bluetooth speaker, the Bluetooth speaker may be woken up while moving, and the position when the control command is issued is different from the position where the wake-up audio is issued. Therefore, this embodiment determines the feedback volume by the audio properties of the control audio. Moreover, after the user issues the control command and then moves, the indoor surveillance camera can be linked or a camera (or sensor) can be set for the Bluetooth speaker to determine the user's position. After determining the feedback information, the feedback volume corresponding to the user's current position is obtained, which can further improve the availability of the feedback volume.

[0054] After verifying the interval time and dialect consistency between the control audio and the wake-up audio, this embodiment identifies the audio attributes of the control audio, calls the attribute-volume curve according to the scene in which the Bluetooth speaker is located, obtains the optimal volume corresponding to the audio attribute, and uses the optimal volume as the feedback volume. In this embodiment, determining the optimal volume by controlling the audio is equivalent to determining the feedback volume according to the user's latest position, which can provide the most appropriate feedback volume for mobile users, and when the feedback volume is obtained using the two parameters of direction and distance, the scene applicability can be improved, ensuring that the feedback volume can meet the user's requirements without the need for the user to adjust frequently.

[0055] Embodiment 4 Based on the third embodiment, this embodiment provides a method for determining the feedback volume in a scenario where multiple control audios are received. In the daily use of Bluetooth speakers, there may be scenarios where multiple people send control commands to the Bluetooth speaker at the same time. For example, after one person sends a wake-up word, multiple people send control commands at the same time. At this time, the Bluetooth speaker will receive multiple control audios. When the audio properties of multiple control audios are different, how to determine the feedback volume? This method is proposed to solve this problem. Please refer to Figure 4 , the specific steps are as follows: When receiving several control audios, the timbre characteristics of each control audio are identified; the timbre characteristics are compared with the timbre characteristics of the preset users, the control audios that match the comparison are used as target audios, and the target audios are analyzed to generate control instructions; wherein the preset users have priority.

[0056] In this embodiment, when the Bluetooth speaker receives multiple control audios, it compares the timbre of each control audio with the wake-up audio, takes the control audio with the same timbre as the target audio, and generates a control instruction based on the target audio, but does not simply determine the feedback volume based on the audio properties of the target audio. It should be noted that in addition to determining the target audio based on the timbre comparison, the target audio can also be determined based on the priority of the user's timbre, that is, analyzing the priority corresponding to the timbre of each control audio (the priority is set in advance), and selecting the highest priority as the target audio.

[0057] While generating control instructions based on the target audio, the feedback volume is determined based on several audio properties of the control audio, including: Identify several audio attributes of the control audio, select the audio attribute closest to the Bluetooth speaker from the several audio attributes as the first target attribute, and use the audio attribute of the target audio as the second target attribute; The volume one is determined according to the target attribute one and the attribute-volume curve, and the volume two is determined according to the target attribute two and the attribute-volume curve; and the feedback volume is determined based on the volume one and the volume two.

[0058] Target attribute 1 corresponds to the user closest to the Bluetooth speaker, and target attribute 2 corresponds to the target user of the Bluetooth speaker feedback information. Of course, it is subject to the target user hearing clearly. If the feedback volume is determined solely based on target attribute 2, the feedback volume may be higher than that of the user corresponding to target attribute 1, which may scare the user.

[0059] Therefore, the feedback volume is determined based on the volume one and the volume two, including: Set volume 1 as the reminder volume and volume 2 as the target volume; After setting the transition between the alert volume and the target volume, the feedback volume is generated.

[0060] Use volume 1 as the reminder volume, and use volume 1 to play reminder words or reminder music. After playing the reminder words or reminder music, use volume 2 to play feedback information, so that the feedback information can be received by the target user while avoiding scaring other users. It should be noted that a transition volume should be set between volume 1 and volume 2 to avoid abrupt jumps between volume 1 and volume 2.

[0061] In some other preferred embodiments, the control audio may be audio emitted by a pet in addition to the relevant control requirements issued by the user, that is, when the pet is close to the Bluetooth speaker, the volume is increased to remind the pet to prevent the pet from going crazy.

[0062] See also Figure 5 The second embodiment of the present application provides a Bluetooth speaker control method based on AI voice analysis, which is applied to a Bluetooth speaker, wherein a data acquisition module and a voice analysis module are in communication and / or electrically connected, including: Data acquisition module: used to receive wake-up audio and control audio; Speech analysis module: used to analyze the wake-up audio and call the audio recognition library, as well as determine the feedback volume according to the audio properties of the control audio and send feedback information according to the feedback volume.

[0063] The sensor is used to collect wake-up audio and control audio, and the audio playback module is used to send feedback information according to the feedback volume. Both the sensor and the audio playback module can be built into the Bluetooth speaker.

[0064] The above embodiments are only used to illustrate the technical method of the present application and are not intended to limit it. Although the present application has been described in detail with reference to the preferred embodiments, a person of ordinary skill in the art should understand that the technical method of the present application may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present application.

Claims

1. A Bluetooth speaker control method based on AI voice analysis, characterized in that: include: Receive a wake-up audio, recognize the wake-up audio and call an audio recognition library; receiving control audio, identifying audio attributes of the control audio, and automatically adjusting the feedback volume of the Bluetooth speaker according to the audio attributes; wherein the audio attributes include direction and distance; and, The control audio is recognized through the audio recognition library, and a control instruction is generated according to the recognition result; the operation is completed according to the control instruction, and feedback information is issued according to the feedback volume.

2. The Bluetooth speaker control method based on AI voice analysis according to claim 1, characterized in that: Before receiving the wake-up audio, setting the wake-up word of the Bluetooth speaker includes: Determine a basic word and several candidate words; combine the basic word and several candidate words in sequence to form a candidate word group; wherein the basic word is a mandatory word set by the Bluetooth speaker manufacturer; An audio test is performed on the plurality of candidate phrases, and at least one candidate phrase is selected as a wake-up word according to the test result; wherein the audio test includes a wake-up rate test, a stability test, and a discrimination rate test.

3. The Bluetooth speaker control method based on AI voice analysis according to claim 2 is characterized in that: Audio testing was performed on several of the candidate phrases, including: Testing the awakening rate and stability of several candidate phrases; The test obtains audio signals of different dialects corresponding to each candidate phrase, and calculates the discrimination rate corresponding to the candidate phrase according to the similarity between the audio signals corresponding to different dialects; Based on the wake-up rate, stability and discrimination rate, the alternative phrase with the best comprehensive performance is selected as the wake-up word.

4. The Bluetooth speaker control method based on AI voice analysis according to claim 1, characterized in that: The identifying and controlling audio attributes of the audio includes: Calculating a time difference between the control audio and the wake-up audio; When the time difference is greater than a set time threshold, a prompt is given to re-awaken the device; otherwise, the relative position of the control audio and the Bluetooth speaker is identified; The audio attribute of the control audio is obtained according to the relative position.

5. The Bluetooth speaker control method based on AI voice analysis according to claim 4 is characterized in that: Automatically adjusting the feedback volume of the Bluetooth speaker according to the audio attribute, including: Retrieving a property-volume curve; wherein the property-volume curve represents a corresponding relationship between audio properties and an optimal volume, and is constructed through simulation testing; The audio attribute is input into the attribute-volume curve, an optimal volume corresponding to the audio attribute is calculated, and the optimal volume is used as the feedback volume.

6. The method for controlling a Bluetooth speaker based on AI voice analysis according to claim 5, characterized in that: Before calculating the time difference between the control audio and the wake-up audio, determining the dialect similarity between the two includes: Determining the dialect similarity between the control audio and the wake-up audio; When the dialect similarity is greater than a similarity threshold, the control audio is recognized by the audio recognition library; otherwise, the control audio is analyzed to re-match the audio recognition library.

7. The method for controlling a Bluetooth speaker based on AI voice analysis according to claim 6, characterized in that: When a number of control audios are received, the timbre characteristics of each control audio are identified; the timbre characteristics are compared with the timbre characteristics of a preset user, the control audio that matches the comparison is used as the target audio, and the target audio is analyzed to generate a control instruction; wherein the preset user has priority.

8. The method for controlling a Bluetooth speaker based on AI voice analysis according to claim 7, characterized in that: The feedback volume is determined according to a number of audio properties of the control audio, including: Identify the audio attributes of the control audio, select the audio attribute closest to the Bluetooth speaker from the audio attributes as target attribute one, and use the audio attribute of the target audio as target attribute two; Volume one is determined according to the target attribute one and the attribute-volume curve, and volume two is determined according to the target attribute two and the attribute-volume curve; feedback volume is determined based on volume one and volume two; wherein volume one is a reminder volume.

9. The Bluetooth speaker control method based on AI voice analysis according to claim 8, characterized in that: Determining the feedback volume based on the volume one and the volume two includes: Using the volume 1 as the reminder volume and the volume 2 as the target volume; After setting a transition between the reminder volume and the target volume, a feedback volume is generated.

10. A Bluetooth speaker control system based on AI voice analysis, applied to a Bluetooth speaker, for executing the Bluetooth speaker control method based on AI voice analysis according to any one of claims 1 to 9, characterized in that: include: Data acquisition module: used to receive wake-up audio and control audio; Speech analysis module: used to analyze the wake-up audio and call the audio recognition library, and determine the feedback volume according to the audio properties of the control audio, and send feedback information according to the feedback volume.

Images