Voice wake-up method, device and system with double protection mechanisms

Abstract

The invention relates to the technical field of voice wake-up, and provides a voice wake-up method, device and system with a dual protection mechanism, and the method comprises the steps: a playing device superposes an interference audio on an original audio to obtain a superposed audio, carries out the wake-up word recognition of the original audio or the superposed audio, obtains a wake-up word interference mark, and carries out the recognition of the wake-up word interference mark; sending the information to a voice service server; the voice wake-up device monitors environment sound, carries out noise processing on the sound carrying interference audio to obtain pure audio, then carries out wake-up word recognition on the pure audio, and generates wake-up word arbitration and sends the wake-up word arbitration to the voice service server when a wake-up word is recognized; the voice service server matches the wake-up word interference mark with the wake-up word arbitration, detects whether wake-up word false triggering exists or not and generates a detection result, and the voice wake-up device executes a wake-up word response or cancels the wake-up word response according to the detection result. Interference of external strong sound source equipment can be reduced, and false triggering of the wake-up word is avoided.

Description

Technical Field

[0001] The present invention relates to the field of voice wake-up technology, and in particular to a voice wake-up method, device, and system with a double protection mechanism. Background Art

[0002] Voice wake-up is a human-computer interaction technology that allows users to activate smart devices by saying a preset wake-up word, so that they can switch from standby or dormant state to working state, ready to receive and execute the next voice command. This technology is widely used in various smart devices such as smartphones, smart speakers, and laptops, improving the convenience and experience of user operation. The main principle of existing voice wake-up technology is to receive voice, recognize wake-up words and voice commands based on microphone arrays to complete corresponding functions.

[0003] Since the accuracy and interaction efficiency of speech recognition are limited or affected by the performance indicators of microphone sensors, environmental noise, and the quality of algorithms. In the prior art, multiple microphones arranged in a specific layout are used to synchronously collect sound, and by analyzing the slight time difference between the signals of each microphone, beamforming and spatial filtering are implemented to construct a directional beam to enhance the speech signal in the target direction while suppressing noise, thereby improving the speech recognition effect. Even in some prior art, the processing effect is dynamically optimized by integrating adaptive algorithms and deep learning technologies to improve speech recognition in noisy environments.

[0004] The prior art has the following technical problems:

[0005] 1. When the background noise is a strong interfering sound source, such as a TV or other audio equipment, the sound source positioning of the voice wake-up device is easily directed to the strong sound source, and the real speaker cannot be identified in time. Especially when playing human voices, the microphone array is easily directed to the TV and mistakenly awakened by keywords in the program, making it difficult to distinguish whether it is a real person speaking or a device playing.

[0006] 2. In a live broadcast scenario, when the presenter speaks the wake-up word, it is easy to wake up a large number of user devices at the same time to access the voice service provider's server, causing a flood attack. Summary of the invention

[0007] The present invention provides a voice wake-up method, device and system with a dual protection mechanism, aiming to solve the technical problems existing in the above-mentioned prior art that when the background noise is a strong interference sound source, it is difficult to distinguish whether it is a real person speaking or a device playing, and when the presenter speaks the wake-up word in the live broadcast scene, it is easy to wake up a large number of user devices at the same time to access the server of the voice service provider, causing a flood attack.

[0008] The technical solution of the present invention to solve the above technical problems is as follows:

[0009] In a first aspect, the present invention provides a voice wake-up method with a dual protection mechanism, which is used in a voice wake-up device, and the method includes:

[0010] Monitoring ambient sound, identifying interference audio from the ambient sound, and performing noise processing on the sound carrying the interference audio to obtain pure audio;

[0011] Performing wake-up word recognition on the clean audio, generating a wake-up word arbitration when a wake-up word is recognized and sending it to the voice service server, wherein the wake-up word arbitration is used to trigger the voice service server to send a detection result to the voice wake-up device after detecting whether the wake-up word is a wake-up word false trigger;

[0012] The detection result is received, and a wake-up word response is executed or a wake-up word response is canceled based on the detection result.

[0013] Furthermore, the interference audio is pre-set in the voice wake-up device, or received from the voice service server.

[0014] Further, the above-mentioned obtaining of pure audio specifically includes:

[0015] Continuously monitor the ambient sound and identify the source direction of various sounds through sound source positioning;

[0016] When interference audio is monitored, the direction of the interference sound source of the interference audio is obtained, the sound source in the direction of the interference sound source is used as noise interference, and the voice wake-up device is used to continue to monitor the ambient sound in other directions to obtain pure audio.

[0017] Furthermore, the above-mentioned wake-up word arbitration includes a second wake-up word identifier and a second timestamp, and the second wake-up word identifier is a wake-up word identifier used to activate the voice wake-up device.

[0018] In a second aspect, the present invention provides a voice wake-up method with a dual protection mechanism, which is used in a voice service server, and the method includes:

[0019] Receive the wake-up word interference mark sent by the playback device and the wake-up word arbitration sent by the voice wake-up device, match the wake-up word interference mark with the wake-up word arbitration, detect whether there is a wake-up word false trigger, and generate a detection result, the wake-up word arbitration is generated when the voice wake-up device recognizes the wake-up word, and the wake-up word interference mark is generated when the voice recognition module built into the playback device recognizes the wake-up word;

[0020] The detection result is sent to the voice wake-up device, where the detection result is used to trigger the voice wake-up device to execute a wake-up word response or cancel the wake-up word response.

[0021] Furthermore, the above also includes:

[0022] Sending interference audio to the playback device, where the interference audio is used to superimpose the original audio of the playback device to obtain superimposed audio;

[0023] An interference audio is sent to a voice wake-up device, where the interference audio is used for filtering the superimposed audio when a voice recognition device recognizes the superimposed audio through encoding of the interference audio.

[0024] Further, the wake-up word interference mark includes a first wake-up word identifier and a first timestamp, the wake-up word arbitration includes a second wake-up word identifier and a second timestamp, and detecting whether there is a wake-up word false trigger specifically includes:

[0025] Determining whether the first wake-up word identifier and the second wake-up word identifier are the same wake-up word;

[0026] If yes, then calculating the time difference between the first timestamp and the second timestamp, and determining whether the time difference is within a statistical confidence interval;

[0027] If so, a detection result indicating that a false trigger exists is generated; if not, a detection result indicating that a false trigger does not exist is generated.

[0028] Furthermore, the calculation formula of the statistical confidence interval is as follows:

[0029]

[0030] Among them, z represents a constant, n represents the number of statistics, and w i represents the observed value of the i-th statistic, Represents the mean of all observations.

[0031] In a third aspect, the present invention provides a voice wake-up method with a dual protection mechanism, which is used in a playback device, and the method includes:

[0032] Superimpose the interference audio onto the original audio to obtain the superimposed audio;

[0033] Performing wake-up word recognition on the original audio or the superimposed audio to obtain a wake-up word interference mark;

[0034] A wake-up word interference mark is sent to the voice service server, and the wake-up word interference mark is used by the voice service server to match the wake-up word interference mark with the wake-up word arbitration, detect whether there is a wake-up word false trigger, and generate a detection result, and the detection result is used to trigger the voice wake-up device to execute a wake-up word response or cancel the wake-up word response. The wake-up word arbitration is generated when the voice wake-up device recognizes the wake-up word.

[0035] Furthermore, the above-mentioned wake-up word interference mark includes a first wake-up word identifier and a first timestamp.

[0036] Furthermore, the interference audio is pre-set in the playback device, or received from the voice service server.

[0037] Furthermore, the above also includes:

[0038] The superimposed audio is played, and the superimposed audio is used to identify and filter the superimposed audio through the interference audio when the voice wake-up device performs noise filtering.

[0039] In a fourth aspect, the present invention provides a voice wake-up device with a dual protection mechanism, which is used in a voice wake-up device, and the device includes:

[0040] A signal processing module is used to monitor the ambient sound, identify the interference audio of the ambient sound, and perform noise processing on the sound carrying the interference audio to obtain pure audio;

[0041] A second wake-up word recognition module is used to recognize the wake-up word of the clean audio, and when the wake-up word is recognized, a wake-up word arbitration is generated and sent to the voice service server, and the wake-up word arbitration is used to trigger the voice service server to send a detection result to the voice wake-up device after detecting whether the wake-up word is a wake-up word false trigger;

[0042] The wake-up word response module is used to receive the detection result and execute the wake-up word response or cancel the wake-up word response based on the detection result.

[0043] In a fifth aspect, the present invention provides a voice wake-up device with a dual protection mechanism, which is used in a voice service server, and the device includes:

[0044] a false trigger detection module, used for receiving a wake-up word interference mark sent by a playback device and a wake-up word arbitration sent by a voice wake-up device, matching the wake-up word interference mark with the wake-up word arbitration, detecting whether there is a false trigger of the wake-up word, and generating a detection result, wherein the wake-up word arbitration is generated when the voice wake-up device recognizes the wake-up word, and the wake-up word interference mark is generated when the voice recognition module built into the playback device recognizes the wake-up word;

[0045] The first sending module is used to send the detection result to the voice wake-up device, and the detection result is used to trigger the voice wake-up device to execute the wake-up word response or cancel the wake-up word response.

[0046] In a sixth aspect, the present invention provides a voice wake-up device with a dual protection mechanism, which is used in a playback device, and the device includes:

[0047] A superposition module, used for superimposing the interference audio onto the original audio to obtain superimposed audio;

[0048] A first wake-up word recognition module, configured to perform wake-up word recognition on the original audio or the superimposed audio to obtain a wake-up word interference mark;

[0049] The second sending module is used to send a wake-up word interference mark to the voice service server, and the wake-up word interference mark is used by the voice service server to match the wake-up word interference mark with the wake-up word arbitration, detect whether there is a false trigger of the wake-up word, and generate a detection result, and the detection result is used to trigger the voice wake-up device to execute the wake-up word response or cancel the wake-up word response. The wake-up word arbitration is generated when the voice wake-up device recognizes the wake-up word.

[0050] In a seventh aspect, the present invention provides a voice wake-up system with a dual protection mechanism, the system comprising a voice wake-up device, a voice service server and a playback device;

[0051] The voice wake-up device includes a voice wake-up device with a dual protection mechanism as described in the fourth aspect, the voice service server includes a voice wake-up device with a dual protection mechanism as described in the fifth aspect, and the playback device includes a voice wake-up device with a dual protection mechanism as described in the sixth aspect. Compared with the prior art, the present invention has the following advantages:

[0052] 1. Effectively reduce the interference of external strong sound source equipment on the microphone array of the voice wake-up device;

[0053] 2. Effectively reduce false trigger interference caused by wake-up words in the audio stream played by external devices;

[0054] 3. The solution has low implementation cost and does not require increasing algorithm complexity or hardware cost.

[0055] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures pointed out in the description, claims and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] In order to more clearly illustrate the embodiments of the present invention 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 some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0057] Figure 1 A schematic flow chart of a voice wake-up method with a dual protection mechanism according to an embodiment of the present invention is shown;

[0058] Figure 2 A schematic flow chart of a voice wake-up method with a dual protection mechanism according to another embodiment of the present invention is shown;

[0059] Figure 3 A schematic diagram of a voice wake-up device according to an embodiment of the present invention is shown;

[0060] Figure 4 A schematic diagram of a voice service server according to an embodiment of the present invention is shown;

[0061] Figure 5 A schematic diagram of a playback device according to an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0062] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0063] The implementation environment involved in each embodiment of the present invention includes: a voice service server, a playback device and a voice wake-up device.

[0064] The voice service server may be a single server, or a server cluster consisting of several servers, or a cloud computing service center.

[0065] The playback devices can be portable music players, smart speakers, Bluetooth speakers, car audio systems, home theater systems, and TVs.

[0066] Voice-activated devices can be smart assistants, smartphones, in-car navigation systems, smart home controllers, and wearable devices.

[0067] Figure 1 FIG. 4 shows a flow chart of a voice wake-up method with a dual protection mechanism according to an embodiment of the present invention. Figure 1 As shown, the voice wake-up method of the embodiment of the present invention includes:

[0068] S1: The playback device recognizes the wake-up word on the original audio, obtains the wake-up word interference mark, and sends the wake-up word interference mark to the voice service server;

[0069] S2: The playback device superimposes the interference audio onto the original audio to obtain superimposed audio;

[0070] S3: The voice wake-up device monitors the ambient sound, identifies the interference audio of the ambient sound, and performs noise processing on the sound carrying the interference audio to obtain pure audio;

[0071] S4: The voice wake-up device recognizes the wake-up word on the clean audio, generates a wake-up word arbitration when a wake-up word is recognized, and sends it to the voice service server;

[0072] S5: The voice service server receives the wake-up word interference mark and the wake-up word arbitration, matches the wake-up word interference mark with the wake-up word arbitration, detects whether there is a wake-up word false trigger, generates a detection result, and sends the detection result to the voice wake-up device;

[0073] S6: The voice wake-up device receives the detection result, and executes a wake-up word response or cancels the wake-up word response based on the detection result.

[0074] Among them, in this embodiment, the voice wake-up device first locates the sound source through a multi-microphone array, and uses the time difference of arrival (TDOA) and intensity difference of the signal to determine the direction of the sound. This process not only helps to focus on the target voice, but also is the basis for identifying the interference source. The interference audio sets a specific frequency coding rule, and the audio signal processing unit in the voice wake-up device identifies the interference audio according to the specific frequency coding rule. When the voice wake-up device detects an audio signal carrying interference audio, it immediately regards the input in the relevant direction as noise interference, and reduces or excludes it in subsequent processing, while giving priority to pure voice signals in other directions. Even if there is noise interference penetrating the front-end defense, a second line of defense is still set in the voice wake-up device. Before confirming the wake-up word response, the voice wake-up device will send an arbitration request to the voice service server to further verify the authenticity of the wake-up event. The voice service server obtains the time difference by comparing the second timestamp with the first timestamp, and analyzes the time difference using a statistical confidence interval. If the two times are closely connected, it indicates that the two recognitions may originate from the same noise event. The voice service server determines this as a false trigger and promptly notifies the voice wake-up device to cancel the wake-up action, avoiding unnecessary system response.

[0075] Optionally, the wake-up word interference mark includes a first wake-up word identifier and a first timestamp, the wake-up word arbitration includes a second wake-up word identifier and a second timestamp, and the voice service server detects whether there is a wake-up word false trigger specifically including:

[0076] Determine whether the first wake-up word identifier and the second wake-up word identifier are the same wake-up word;

[0077] If yes, then calculating the time difference between the first timestamp and the second timestamp, and determining whether the time difference is within a statistical confidence interval;

[0078] If yes, a detection result indicating false triggering is generated; if no, a detection result indicating no false triggering is generated;

[0079] Optionally, the calculation method of the statistical confidence interval includes: a confidence interval based on a normal distribution, a confidence interval based on a t distribution, and a confidence interval of a proportion;

[0080] Among them, when the number of statistics is large enough (n>30), the statistical confidence interval adopts the confidence interval based on the normal distribution, and the calculation formula of the statistical confidence interval is shown as follows:

[0081]

[0082] Among them, z represents a constant, n represents the number of statistics, and w i represents the observed value of the i-th statistic, Represents the mean of all observations.

[0083] Among them, when the number of statistics is small (n≤30), the statistical confidence interval adopts the confidence interval based on t distribution;

[0084] Since the detection results are of two types: the time difference is within the statistical confidence interval and the time difference is not within the statistical confidence interval, the statistical confidence interval can also use a proportional confidence interval.

[0085] Optionally, the interference audio is pre-set in the voice wake-up device, or is received from the voice service server.

[0086] Optionally, the interference audio is pre-set in the playback device, or received from the voice service server.

[0087] Optionally, there are multiple encoding settings for the interference audio, and different playback devices receive different encodings of the interference audio; or different interference audios are preset in different playback devices;

[0088] The voice wake-up device receives the codes of all interfering audios and can identify all interfering audios; or the codes of all interfering audios are built into the voice wake-up device.

[0089] Optionally, in S1, at the playback device end, the currently received data stream is first identified as audio content through a built-in hardware or software mechanism. At this stage, the playback device will check the format, encoding type and other information of the data packet to ensure the accuracy of subsequent processing. After confirming that it is an audio stream, the decoder in the playback device is started, which is responsible for decoding the compressed audio data (such as MP3, AAC and other formats) into a format that can be directly processed by the first speech recognition module, such as an uncompressed PCM (pulse code modulation) format.

[0090] Optionally, in S2, a section of interference audio with a specific frequency coding rule is superimposed on the decoded data. The frequency coding rule can be implemented at a low frequency or high frequency end to which the human ear is insensitive. For example, three frequencies of 16.5KHz, 17KHz, and 17.5KHz are recorded as A, B, and C, and an AC-B style encoded sound wave signal is generated every t milliseconds. Among them, the AC-B style encoded sound wave signal is only an encoding of one interference audio. The encoding of n types of interference audio can be distinguished by swapping the order of A, B, and C, such as AB-C, BC-A and other combination encodings, and the frequency D can also be increased to increase the combination encoding of the interference audio.

[0091] In order to identify and manage the interference sound source later, a section of interference audio encoded with a specific frequency is superimposed on the decoded audio stream. This special interference sound wave signal will not affect the user experience, but it is a recognizable mark for the voice wake-up device to track and analyze the interference in the audio stream.

[0092] Optional, such as Figure 2 As shown, S1 and S2 can be exchanged in order. The playback device first superimposes the interference audio on the original audio through the superposition module, and then continuously monitors the superimposed audio output by the superposition module to find the preset wake-up word. When the wake-up word is recognized, a mark, namely the wake-up word interference mark, is sent to the voice service server.

[0093] After receiving the wake-up word interference mark, the voice service server records it in a log or database. This step is useful for subsequent wake-up word false triggering judgment.

[0094] Optionally, in S3, based on the frequency coding rules of the interference audio, the voice wake-up device is used to analyze the input signal and perform the sound source localization function. The direction of the sound source is identified by comparing the audio responses from different directions. During this process, when the voice wake-up device detects the coding frequency of the interference audio (such as the frequency of the AC-B-AC-B or B-AC-B-AC style coding mentioned above) within 5t consecutive milliseconds, it is considered to be interference audio, and the sound source in this direction is automatically determined to be noise interference. The signal from this direction is temporarily ignored, and the focus is instead on monitoring the pure audio input from other directions.

[0095] Among them, this embodiment uses the preset frequency coding rules to analyze the input signal to detect the interference audio. And through the sound source localization function, the signal from the interference audio direction is ignored, and the focus is more on monitoring the pure voice input from other directions to improve the accuracy of wake-up word recognition.

[0096] Optionally, in S4, when the interference audio breaks through the front-end defense line and enters the wake-up word response processing flow, the voice wake-up device still has a set of redundant mechanisms to prevent the wake-up word from being triggered by mistake. Before the voice wake-up device recognizes the wake-up word and responds to the wake-up word event, the voice wake-up device will immediately send a wake-up word arbitration to the voice service server.

[0097] In this embodiment, the wake-up word arbitration is sent to the voice service server to start an additional verification process to avoid false activation caused by noise.

[0098] Optionally, in S5, the calculation formula for the statistical confidence interval is:

[0099] Statistical confidence interval = sample mean + (critical value × standard error)

[0100] When the sample size is large enough (greater than 30 in this embodiment), the critical value can be calculated based on the normal distribution function and the confidence level of 95%, and the value is 1.96. Assume that the sample size is n and the sample mean is If the sample standard deviation is s, then The specific formula for obtaining the statistical confidence interval is:

[0101]

[0102] Among them, w i represents the observed value of the i-th sample, represents the mean of all observations, and n represents the number of observations in the sample.

[0103] Among them, in this embodiment, after the voice service server receives the wake-up word arbitration sent by the voice wake-up device, it will start a verification process, compare the previously recorded first timestamp with the newly received second timestamp, and calculate the time difference between the two. And compare the obtained time difference with the pre-set statistical confidence interval. If the time difference falls within the statistical confidence interval, it means that the two recognized wake-up words are very likely to come from the same event, and the second recognition may be a misjudgment caused by noise interference. In this case, a detection result with false triggering is generated; otherwise, a detection result without false triggering is generated.

[0104] like Figure 2As shown, it shows a structural schematic diagram of a voice wake-up device with a dual protection mechanism provided by an embodiment of the present invention, and the voice wake-up device can be applied to a voice wake-up device, including:

[0105] A signal processing module is used to monitor the ambient sound, identify the interference audio of the ambient sound, and perform noise processing on the sound carrying the interference audio to obtain pure audio;

[0106] A second wake-up word recognition module is used to recognize the wake-up word of the clean audio, and when the wake-up word is recognized, a wake-up word arbitration is generated and sent to the voice service server, and the wake-up word arbitration is used to trigger the voice service server to send a detection result to the voice wake-up device after detecting whether the wake-up word is a wake-up word false trigger;

[0107] The wake-up word response module is used to receive the detection result and execute the wake-up word response or cancel the wake-up word response based on the detection result.

[0108] like Figure 3 As shown, it shows a voice wake-up device with a dual protection mechanism provided by an embodiment of the present invention, which is used in a voice service server, and the device includes:

[0109] a false trigger detection module, used for receiving a wake-up word interference mark sent by a playback device and a wake-up word arbitration sent by a voice wake-up device, matching the wake-up word interference mark with the wake-up word arbitration, detecting whether there is a false trigger of the wake-up word, and generating a detection result, wherein the wake-up word arbitration is generated when the voice wake-up device recognizes the wake-up word, and the wake-up word interference mark is generated when the voice recognition module built into the playback device recognizes the wake-up word;

[0110] The first sending module is used to send the detection result to the voice wake-up device, and the detection result is used to trigger the voice wake-up device to execute the wake-up word response or cancel the wake-up word response.

[0111] like Figure 4 As shown, it shows a voice wake-up device with a dual protection mechanism provided by an embodiment of the present invention, which is used in a playback device, and the device includes:

[0112] A superposition module, used for superimposing the interference audio onto the original audio to obtain superimposed audio;

[0113] A first wake-up word recognition module, configured to perform wake-up word recognition on the original audio or the superimposed audio to obtain a wake-up word interference mark;

[0114] The second sending module is used to send a wake-up word interference mark to the voice service server, and the wake-up word interference mark is used by the voice service server to match the wake-up word interference mark with the wake-up word arbitration, detect whether there is a false trigger of the wake-up word, and generate a detection result, and the detection result is used to trigger the voice wake-up device to execute the wake-up word response or cancel the wake-up word response. The wake-up word arbitration is generated when the voice wake-up device recognizes the wake-up word.

[0115] The present invention provides a voice wake-up system with a dual protection mechanism, the system comprising a voice wake-up device, a voice service server and a playback device;

[0116] A playback device, configured to perform wake-up word recognition on the original audio, obtain a wake-up word interference mark, send the wake-up word interference mark to the voice service server, and superimpose the interference audio onto the original audio to obtain superimposed audio;

[0117] A voice wake-up device is used to monitor ambient sounds, identify interference audio from the ambient sounds, perform noise processing on the sound carrying the interference audio to obtain pure audio, perform wake-up word recognition on the pure audio, and generate a wake-up word arbitration when a wake-up word is recognized and send it to the voice service server;

[0118] A voice service server, used to match the wake-up word interference mark with the wake-up word arbitration, detect whether there is a wake-up word false trigger, generate a detection result, and send the detection result to the voice wake-up device;

[0119] The voice wake-up device is also used to receive the detection result and execute the wake-up word response or cancel the wake-up word response based on the detection result.

[0120] Optionally, the playback device is also used to replace the original audio with the superimposed audio to perform wake-up word recognition and obtain a wake-up word interference mark.

[0121] Optionally, also include:

[0122] The switch module is used to construct a global voice protection switch. When the global voice protection switch is in the on state, the voice wake-up system with a dual protection mechanism is enabled.

[0123] The above description is only a preferred embodiment of the present invention and an explanation of the technical principles used. Those skilled in the art should understand that the disclosure scope involved in the present invention is not limited to the technical solution formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above disclosed concept. For example, the above features are replaced with the technical features with similar functions disclosed in the present invention (but not limited to) by each other.

Claims

1. A voice wake-up method with a dual protection mechanism, characterized in that: Used in a voice wake-up device, the method includes: Monitor the ambient sound, identify the interference audio of the ambient sound, and perform noise processing on the sound carrying the interference audio to obtain pure audio; Performing wake-up word recognition on the clean audio, generating a wake-up word arbitration when a wake-up word is recognized and sending it to the voice service server, wherein the wake-up word arbitration is used to trigger the voice service server to send a detection result to the voice wake-up device after detecting whether the wake-up word is a wake-up word false trigger; The detection result is received, and a wake-up word response is executed or a wake-up word response is canceled based on the detection result.

2. According to claim 1, a voice wake-up method with a dual protection mechanism is characterized in that: The interference audio is pre-set in the voice wake-up device, or received from the voice service server.

3. According to claim 1, a voice wake-up method with a dual protection mechanism is characterized in that: Getting pure audio specifically includes: Continuously monitor the ambient sound and identify the source direction of various sounds through sound source positioning; When interference audio is monitored, the direction of the interference sound source of the interference audio is obtained, the sound source in the direction of the interference sound source is used as noise interference, and the voice wake-up device is used to continue to monitor the ambient sound in other directions to obtain pure audio.

4. According to claim 1, a voice wake-up method with a dual protection mechanism is characterized in that: The wake-up word arbitration includes a second wake-up word identifier and a second timestamp, where the second wake-up word identifier is a wake-up word identifier used to activate a voice wake-up device.

5. A voice wake-up method with a dual protection mechanism, characterized in that: Used in a voice service server, the method comprises: Receive the wake-up word interference mark sent by the playback device and the wake-up word arbitration sent by the voice wake-up device, match the wake-up word interference mark with the wake-up word arbitration, detect whether there is a wake-up word false trigger, and generate a detection result, the wake-up word arbitration is generated when the voice wake-up device recognizes the wake-up word, and the wake-up word interference mark is generated when the voice recognition module built into the playback device recognizes the wake-up word; The detection result is sent to the voice wake-up device, where the detection result is used to trigger the voice wake-up device to execute a wake-up word response or cancel the wake-up word response.

6. According to claim 5, a voice wake-up method with a double protection mechanism is characterized in that: Also includes: Sending interference audio to the playback device, where the interference audio is used to superimpose the original audio of the playback device to obtain superimposed audio; An interference audio is sent to a voice wake-up device, where the interference audio is used for filtering the superimposed audio when a voice recognition device recognizes the superimposed audio through encoding of the interference audio.

7. According to claim 5, a voice wake-up method with a double protection mechanism is characterized in that: The wake-up word interference mark includes a first wake-up word identifier and a first timestamp, the wake-up word arbitration includes a second wake-up word identifier and a second timestamp, and detecting whether there is a wake-up word false trigger specifically includes: Determine whether the first wake-up word identifier and the second wake-up word identifier are the same wake-up word; if so, calculate the time difference between the first timestamp and the second timestamp, and determine whether the time difference is within a statistical confidence interval; If so, a detection result indicating that a false trigger exists is generated; if not, a detection result indicating that a false trigger does not exist is generated.

8. A voice wake-up method with a dual protection mechanism according to claim 7, characterized in that: The calculation formula of the statistical confidence interval is as follows: Among them, z represents a constant, n represents the number of statistics, and w i represents the observed value of the i-th statistic, Represents the mean of all observations.

9. A voice wake-up method with a dual protection mechanism, characterized in that: Used in a playback device, the method comprises: Superimpose the interference audio onto the original audio to obtain the superimposed audio; Performing wake-up word recognition on the original audio or the superimposed audio to obtain a wake-up word interference mark; A wake-up word interference mark is sent to the voice service server, and the wake-up word interference mark is used by the voice service server to match the wake-up word interference mark with the wake-up word arbitration, detect whether there is a wake-up word false trigger, and generate a detection result, and the detection result is used to trigger the voice wake-up device to execute a wake-up word response or cancel the wake-up word response. The wake-up word arbitration is generated when the voice wake-up device recognizes the wake-up word.

10. A voice wake-up method with a double protection mechanism according to claim 9, characterized in that: The wake-up word interference mark includes a first wake-up word identifier and a first timestamp.

11. A voice wake-up method with a double protection mechanism according to claim 9, characterized in that: The interference audio is pre-set in the playback device, or received from the voice service server.

12. A voice wake-up method with a double protection mechanism according to claim 11, characterized in that: Also includes: The superimposed audio is played, and the superimposed audio is used to identify and filter the superimposed audio through the interference audio when the voice wake-up device performs noise filtering.

13. A voice wake-up device with a dual protection mechanism, characterized in that: Used in a voice wake-up device, the device comprises: A signal processing module is used to monitor the ambient sound, identify the interference audio of the ambient sound, and perform noise processing on the sound carrying the interference audio to obtain pure audio; A second wake-up word recognition module is used to recognize the wake-up word of the clean audio, and when the wake-up word is recognized, a wake-up word arbitration is generated and sent to the voice service server, and the wake-up word arbitration is used to trigger the voice service server to send a detection result to the voice wake-up device after detecting whether the wake-up word is a wake-up word false trigger; The wake-up word response module is used to receive the detection result and execute the wake-up word response or cancel the wake-up word response based on the detection result.

14. A voice wake-up device with a dual protection mechanism, characterized in that: Used in a voice service server, the device comprises: a false trigger detection module, used for receiving a wake-up word interference mark sent by a playback device and a wake-up word arbitration sent by a voice wake-up device, matching the wake-up word interference mark with the wake-up word arbitration, detecting whether there is a false trigger of the wake-up word, and generating a detection result, wherein the wake-up word arbitration is generated when the voice wake-up device recognizes the wake-up word, and the wake-up word interference mark is generated when the voice recognition module built into the playback device recognizes the wake-up word; The first sending module is used to send the detection result to the voice wake-up device, and the detection result is used to trigger the voice wake-up device to execute the wake-up word response or cancel the wake-up word response.

15. A voice wake-up device with a dual protection mechanism, characterized in that: Used in a playback device, the device comprises: A superposition module, used for superimposing the interference audio onto the original audio to obtain superimposed audio; A first wake-up word recognition module, configured to perform wake-up word recognition on the original audio or the superimposed audio to obtain a wake-up word interference mark; The second sending module is used to send a wake-up word interference mark to the voice service server, and the wake-up word interference mark is used by the voice service server to match the wake-up word interference mark with the wake-up word arbitration, detect whether there is a false trigger of the wake-up word, and generate a detection result, and the detection result is used to trigger the voice wake-up device to execute the wake-up word response or cancel the wake-up word response. The wake-up word arbitration is generated when the voice wake-up device recognizes the wake-up word.

16. A voice wake-up system with a dual protection mechanism, characterized in that: The system includes a voice wake-up device, a voice service server and a playback device; The voice wake-up device includes a voice wake-up device with a double protection mechanism as described in claim 13, the voice service server includes a voice wake-up device with a double protection mechanism as described in claim 14, and the playback device includes a voice wake-up device with a double protection mechanism as described in claim 15.

Images