Distributed speech recognition device wake-up method, terminal and storage medium

Abstract

The present invention relates to a distributed voice recognition device wake-up method, terminal and storage medium, which belongs to the field of voice recognition, wherein the method includes: a central device or a non-central device identifies the received voice wake-up information and generates a recognition result corresponding to the device itself; the non-central device sends the recognition result obtained by its own processing to the central device; the central device processes and obtains the actual information to be executed based on all the recognition results; the central device identifies the target recognition result with the largest signal strength value, and marks the central device or non-central device that generates the target recognition result as a response device; when the response device is a central device, the central device completes the response based on the actual information to be executed; when the response device is a non-central device, the central device sends the actual information to be executed to the response device to complete the response. The present invention has the effect of making the response timing of the voice recognition device more accurate and improving the user experience.

Description

Technical Field

[0001] The present invention relates to the field of speech recognition, and in particular to a distributed speech recognition device wake-up method, a terminal and a storage medium. Background Art

[0002] Speech recognition technology, also known as automatic speech recognition, aims to convert the vocabulary content in human speech into computer-readable input, such as keystrokes, binary codes, or character sequences.

[0003] Currently, there are more and more devices with voice recognition functions, such as smartphones, watches, smart home groups, etc. The use scenarios of voice interaction control are also increasing.

[0004] In the process of implementing the present application, the inventors found that the above technology has at least the following problems: in daily use, there may be multiple devices with voice recognition function in the same area at the same time. When the user wants to wake up one of the devices, the voice emitted by the user will be obtained by multiple devices at the same time, resulting in multiple devices being woken up and responding at the same time, which is noisy and reduces the user experience. Summary of the invention

[0005] In order to make the response timing of the voice recognition device more accurate and improve the user experience, the present application provides a distributed voice recognition device wake-up method, terminal and storage medium.

[0006] In a first aspect, the present application provides a distributed speech recognition device wake-up method, which adopts the following technical solution:

[0007] A distributed voice recognition device wake-up method, the method is applied to a distributed collaborative device group consisting of at least two voice recognition devices, the voice recognition devices are divided into a central device and at least one non-central device, the method comprising:

[0008] The hub device or non-hub device identifies the received voice wake-up information and generates a recognition result corresponding to the device itself, wherein the recognition result carries a signal strength value of the corresponding voice wake-up information;

[0009] The non-central device sends the recognition result obtained by its own processing to the central device;

[0010] The central device processes and obtains actual information to be executed based on all the recognition results;

[0011] The hub device identifies the target identification result carrying the largest signal strength value, and marks the hub device or non-hub device that generates the target identification result as a responding device;

[0012] When the responding device is a hub device, the hub device completes the response based on the actual information to be executed;

[0013] When the responding device is a non-central device, the central device sends the actual information to be executed to the responding device to enable it to complete the response.

[0014] By adopting the above technical solution, users can select one of the voice recognition devices in the distributed collaborative device group as a central device based on actual needs, and the remaining devices are non-central devices. In daily use, when the user verbally expresses his needs, the voice recognition devices within a certain range will take over the voice wake-up information and identify the signal strength value of the voice wake-up information received by itself. After that, the recognition results of each voice recognition device will be summarized to the central device, and the central device will generate actual information to be executed based on the recognition results. At the same time, based on the signal strength value, a response device is selected from all voice recognition devices. After that, the response device completes the response based on the actual information to be executed, so that the response timing of the voice recognition device is more accurate, reducing the possibility that all devices that receive the voice wake-up information will respond, and improving the user experience.

[0015] Optionally, each of the voice recognition devices stores a voice recognition library corresponding to itself, and the voice recognition library records the correspondence between the wake-up word and the executable operation corresponding to the voice recognition device;

[0016] Before the hub device or the non-hub device recognizes the received voice wake-up information and generates a recognition result corresponding to the device itself, it includes:

[0017] After receiving the voice information, the hub device or the non-hub device identifies whether there is a wake-up word in the voice information based on the corresponding voice recognition library;

[0018] When the hub device or the non-hub device recognizes that the voice information contains a wake-up word corresponding to itself, it authenticates the voice information as voice wake-up information.

[0019] By adopting the above technical solution, after receiving the voice information, the voice recognition device will verify the voice information based on its own stored voice recognition library, thereby filtering out the voice information that does not contain the wake-up word and reducing unnecessary processing.

[0020] Optionally, the recognition result also carries a wake-up word recognized from the voice wake-up information;

[0021] The central device processes all the recognition results to obtain actual information to be executed, which specifically includes:

[0022] The central device compares the signal strength value carried in each recognition result with a preset strength threshold, and marks the recognition result carrying a signal strength value greater than the strength threshold as a valid recognition result;

[0023] The central device counts the wake-up words carried by all valid recognition results to identify the wake-up word with the highest frequency of occurrence;

[0024] The central device generates actual information to be executed based on the identified wake-up word with the highest occurrence frequency.

[0025] By adopting the above technical solution, when a voice recognition device is far away from the sound source, the accuracy of the voice wake-up information it receives may also be affected. Therefore, the central device only marks the recognition results carrying signal strength values ​​greater than the strength threshold as valid recognition results. After that, the central device identifies the most frequently occurring wake-up words from all recognition results, and further filters out occasional incorrectly recognized wake-up words, thereby helping to further improve the accuracy of the processing results.

[0026] Optionally, the speech recognition device stores correspondences between different speech recognition devices and device-pointing nouns;

[0027] The method further comprises:

[0028] When the non-central device recognizes that there is a device-directed noun in the received voice information, it cancels the recognition of the wake-up word in the voice information and sends the voice information carrying the device-directed noun to the central device;

[0029] After receiving the voice information carrying the device-pointing noun sent by the user or the non-hub device, the hub device identifies the target voice recognition device corresponding to the device-pointing noun;

[0030] When the hub device recognizes that the target speech recognition device is a hub device, the hub device recognizes and processes the speech information carrying the device-pointing noun, and completes the response after recognizing the wake-up word corresponding to itself;

[0031] When the central device identifies that the target voice recognition device is a non-central device, it sends the voice information carrying the device-pointing noun to the target voice recognition device, so that the target voice recognition device recognizes and processes the voice information, and enables the target voice recognition device to complete the response after recognizing the wake-up word corresponding to itself.

[0032] By adopting the above technical solution, on the one hand, the user can specify the device he wants to control by directly speaking the device pointing noun, thereby improving processing efficiency and the accuracy of response device selection; on the other hand, even if the user's voice is not directly received by the target voice recognition device pointed to by the device pointing noun, as long as a central device or a non-central device receives the voice information carrying the device pointing noun, it can forward the corresponding voice information to the target voice recognition device, thereby expanding the effective control range, allowing users to voice control voice recognition devices at a longer distance, further improving the user experience.

[0033] Optionally, if the central device receives multiple voice messages carrying the device-pointing noun within a preset recognition time, the central device only performs further recognition processing on the first received voice message carrying the device-pointing noun.

[0034] By adopting the above technical solution, when a user sends a voice message carrying a device-pointing noun and it is received by multiple voice devices at the same time, in order to avoid repeated processing of the same voice message, the central device will only perform further recognition processing on the first received voice message carrying the device-pointing noun.

[0035] Optionally, the method further includes:

[0036] After receiving the speech recognition library update instruction, the central device or non-central device updates the currently stored speech recognition library based on the speech recognition library update instruction, and the speech recognition library update instruction records the correspondence between the new wake-up word and the executable operation.

[0037] By adopting the above technical solution, users can update the speech recognition library stored in the speech recognition device according to personal habits and actual application scenarios.

[0038] Optionally, before the hub device identifies the target identification result carrying the largest signal strength value and marks the hub device or non-hub device that generates the target identification result as a responding device, the method further includes:

[0039] The central device identifies whether there is a voice recognition device currently in an active state, and the voice recognition device in an active state refers to the voice recognition device that last responded within a preset historical time period;

[0040] When the recognition result is yes, the hub device cancels the comparison of the signal strength value and marks the active voice recognition device as a responding device.

[0041] By adopting the above technical solution, when the central device recognizes the existence of an active device, it can directly mark the active device as a responding device without comparing the signal strength value, which helps to improve processing efficiency.

[0042] In a second aspect, the present application provides a speech recognition device, which adopts the following technical solution:

[0043] A speech recognition device, which can work as a hub device or a non-hub device based on user selection, and includes:

[0044] A voice information receiving module, used to receive the voice wake-up information sent by the user;

[0045] An information recognition module, used to recognize the received voice wake-up information and generate a recognition result corresponding to the device itself;

[0046] An information interaction module, used for information interaction with other speech recognition devices;

[0047] An actual information to be executed generation module, used for processing and obtaining the actual information to be executed based on all the recognition results;

[0048] A responding device marking module, used to identify the target recognition result carrying the maximum signal strength value, and mark the hub device or non-hub device that generates the target recognition result as a responding device;

[0049] The response execution module is used to complete the response based on the actual information to be executed.

[0050] In a third aspect, the present application provides a smart terminal, which adopts the following technical solution:

[0051] An intelligent terminal comprises a memory and a processor, wherein the memory stores a computer program that can be loaded by the processor and execute the method as described in the first aspect.

[0052] In a fourth aspect, the present application provides a computer-readable storage medium, which adopts the following technical solution:

[0053] A computer-readable storage medium includes a computer program that can be loaded by a processor and execute the method as described in the first aspect.

[0054] In summary, the present application includes at least one of the following beneficial technical effects:

[0055] 1. Based on actual needs, users select one of the voice recognition devices in the distributed collaborative device group as the central device, and the remaining devices as non-central devices. In daily use, when the user verbally expresses his or her needs, the voice recognition devices within a certain range will receive the voice wake-up information and identify the signal strength value of the voice wake-up information received by itself. After that, the recognition results of each voice recognition device will be summarized to the central device, and the central device will generate actual information to be executed based on the recognition results. At the same time, based on the signal strength value, the responding device will be selected from all voice recognition devices. After that, the responding device will complete the response based on the actual information to be executed, so that the response timing of the voice recognition device is more accurate, reducing the possibility that all devices that receive the voice wake-up information will respond, and improving the user experience;

[0056] 2. After receiving the voice information, the voice recognition device will verify the voice information based on its own stored voice recognition library, thereby filtering out the voice information that does not contain the wake-up word and reducing unnecessary processing;

[0057] 3. Through the recognition and processing of device-pointing nouns, on the one hand, the user can specify the device he wants to control by directly speaking the device-pointing noun, thereby improving processing efficiency and the accuracy of response device selection; on the other hand, even if the user's voice is not directly received by the target voice recognition device pointed to by the device-pointing noun, as long as a central device or a non-central device receives the voice information carrying the device-pointing noun, it can forward the corresponding voice information to the target voice recognition device, thereby expanding the effective control range, allowing users to voice control voice recognition devices at a longer distance, and further improving the user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 is a schematic diagram for reflecting a distributed collaborative device group composed of multiple speech recognition devices in an embodiment of the present application;

[0059] Figure 2 It is a flowchart for embodying a distributed speech recognition device wake-up method in an embodiment of the present application;

[0060] Figure 3 It is a system block diagram used to embody the speech recognition device in the embodiment of the present application.

[0061] Explanation of the reference numerals: 31. Voice information receiving module; 32. Information identification module; 33. Information interaction module; 34. Actual information to be executed generating module; 35. Response device marking module; 36. Response execution module. DETAILED DESCRIPTION

[0062] The following is combined with Figure 1-3This application is described in further detail.

[0063] The present application embodiment discloses a distributed voice recognition device wake-up method, referring to Figure 1 The method can be applied to a distributed collaborative device group consisting of at least two speech recognition devices, and the speech recognition devices belonging to the same distributed collaborative device group can be located in the same spatial area. Each speech recognition device includes a MIC array for collecting speech information, an AI processing chip for information processing, a storage chip for data storage, and a communication component for signal interaction with other speech recognition devices, wherein the communication component can be a Bluetooth module, a wireless network card, etc.

[0064] Among the voice recognition devices belonging to the same distributed collaborative device group, there is a central device, and the remaining voice recognition devices are non-central devices. The central device can be selected by the user according to actual needs. Generally, the central devices are voice recognition devices that are in operation for a long time. For example, when the application environment is in the user's home, the voice recognition device can be a refrigerator, air conditioner, table lamp and other smart devices with voice recognition function. Since the refrigerator is in operation under normal circumstances, the user can select the refrigerator as the central device.

[0065] The following will be combined with specific implementation methods. Figure 2 The processing flow shown is described in detail, and the content can be as follows:

[0066] S201: The hub device or the non-hub device identifies the received voice wake-up information and generates a recognition result corresponding to the device itself.

[0067] Among them, the recognition result carries the signal strength value of the corresponding voice wake-up information. The signal strength value can be reflected by the sound pressure of the sound. Generally speaking, since the sound waves will gradually attenuate as the transmission distance increases during the transmission process, the farther the voice recognition device is from the sound source, the lower the signal strength value of the received voice wake-up information.

[0068] In implementation, when the user needs to control the voice recognition device, he can speak out his needs. After the user's voice is transmitted through the air, it will be acquired by the voice recognition device in the nearby range, that is, the central device or non-central device will receive the voice wake-up information. After receiving the voice wake-up information, the central device or non-central device will perform voice recognition, strength recognition and other recognition processing on the voice wake-up information, so as to confirm the signal strength value of the voice wake-up information received by itself, and generate a recognition result carrying the voice recognition result and the signal strength value.

[0069] S202: The non-central device sends the recognition result obtained by its own processing to the central device.

[0070] In implementation, when the voice recognition device that receives the voice wake-up information is a non-central device, the non-central device will send the recognition result obtained by its own processing to the central device.

[0071] S203: The central device processes and obtains actual information to be executed based on all recognition results.

[0072] The actual information to be executed is information that can be processed by a voice recognition device to obtain actual execution instructions.

[0073] In implementation, the hub device processes the information to be executed based on the speech recognition results among all the recognition results obtained within the preset waiting time. The preset waiting time can be pre-set by the staff according to the size of the room or area, so as to reserve enough time for the propagation of sound and the transmission of signals. The recognition result can be obtained by the hub device itself or received by the hub device from a non-hub device.

[0074] The central device starts timing after receiving the first recognition result, and stops receiving the recognition results of this round after the preset waiting time.

[0075] S204: The hub device identifies the target identification result carrying the largest signal strength value, and marks the hub device or non-hub device that generates the target identification result as a responding device.

[0076] In implementation, the hub device identifies and compares the signal strength values ​​carried by all the received recognition results, thereby confirming the target recognition result with the largest signal strength value. Afterwards, the hub device marks the speech recognition device that generates the target recognition result as a responding device, where the responding device can be the hub device itself or a non-hub device.

[0077] S205: When the responding device is a central device, the central device completes the response based on the actual information to be executed.

[0078] In implementation, when the responding device marked by the central device is the central device itself, the central device processes the actual information to be executed to obtain actual instructions to be executed that can be executed by itself. Thereafter, the central device executes the above actual instructions to be executed to complete the response.

[0079] S206: When the responding device is a non-central device, the central device sends the actual information to be executed to the responding device to enable it to complete the response.

[0080] In implementation, when the responding device marked by the central device is a non-central device, the central device sends the actual information to be executed to the non-central device marked as the responding device. After receiving the actual information to be executed, the non-central device processes the actual information to be executed to obtain actual instructions to be executed that can be executed by itself. Thereafter, the non-central device executes the above actual instructions to be executed to complete the response.

[0081] Optionally, since the user's speech is not necessarily to issue instructions to the voice recognition device, in order to filter out such sounds, in another embodiment, each voice recognition device stores a voice recognition library corresponding to itself, and each voice recognition library records the corresponding relationship between the wake-up word and the executable operation corresponding to the voice recognition device. In this case, before the above S201, the following processing may also be included:

[0082] After receiving the voice information, the hub device or non-hub device identifies whether there is a wake-up word in the voice information based on its corresponding voice recognition library.

[0083] In practice, the user's speech can be received by the voice recognition device after being transmitted through the air, that is, the central device or non-central device receives the voice information. Afterwards, the central device or non-central device that receives the voice information can perform noise reduction processing on the received voice device, and compare the voice information after noise reduction processing with the voice recognition library stored in itself to identify whether the above voice information contains the corresponding wake-up word recorded in the voice recognition library.

[0084] When the hub device or the non-hub device recognizes that the received voice information contains a wake-up word corresponding to itself, it authenticates the voice information as voice wake-up information.

[0085] In implementation, for a hub device or any non-hub device, on the one hand, when it is recognized that there is a wake-up word in the received voice information, it means that the device can perform corresponding executable operations based on the voice information. At this time, the device will authenticate the voice information it received as voice wake-up information. On the other hand, when it is recognized that there is no wake-up word in the received voice information, the device will cancel further processing of the received voice information.

[0086] Furthermore, in another embodiment, the recognition result generated in the above S201 also carries the wake-up word recognized from the voice information authenticated as the voice wake-up information. In this case, the above S203 may specifically include the following contents:

[0087] In the first step, the central device compares the signal strength value carried in each recognition result with the preset strength threshold, and marks the recognition result carrying the signal strength value greater than the above strength threshold as a valid recognition result.

[0088] The intensity threshold may be preset by the user or may be set when the speech recognition device leaves the factory.

[0089] In implementation, for each recognition result obtained, the central device will compare the signal strength value carried by it with the preset strength threshold. Generally speaking, when a voice recognition device is far away from the sound source, the accuracy of the voice wake-up information it receives may also be affected. Therefore, for the recognition results with signal strength values ​​lower than the strength threshold, the central device can ignore them and only mark the recognition results with signal strength values ​​greater than the strength threshold as valid recognition results, which helps to improve the accuracy of the processing results for further processing.

[0090] In the second step, the central device counts the wake-up words carried by all valid recognition results to identify the wake-up words with the highest frequency.

[0091] During implementation, the central device counts the number of times each wake-up word appears in all valid recognition results, and identifies the wake-up word with the highest frequency, thereby filtering out occasional, misrecognized wake-up words.

[0092] In the third step, the central device generates the actual information to be executed based on the wake-up word with the highest frequency of occurrence.

[0093] In implementation, after the central device identifies the wake-up word with the highest frequency, it can generate actual information to be executed that carries the wake-up word with the highest frequency. In this case, when the above actual information to be executed is sent to the responding device, the responding device will perform recognition processing on it, thereby identifying the wake-up word in the actual information to be executed, and then based on the wake-up word, identify the corresponding executable operation from its own pre-stored voice recognition library and execute it, thereby completing the response.

[0094] Optionally, in another embodiment, the voice recognition device may also store correspondences between different voice recognition devices and device-pointing nouns, where the device-pointing noun may be a device name, a device name carrying location information, etc. For example, the device-pointing noun may be a "refrigerator" corresponding to a refrigerator, a "bedroom air conditioner" corresponding to an air conditioner installed in a bedroom, etc., thereby facilitating the user to verbally specify the voice recognition device they want to control.

[0095] In this case, the following may also be included:

[0096] When the non-central device recognizes that there is a device-directed noun in the received voice information, it cancels the recognition of the wake-up word in the voice information and sends the voice information carrying the device-directed noun to the central device.

[0097] In implementation, after receiving voice information, the non-hub device will identify whether there is a device-pointing noun in the voice information. When it is identified that there is a device-pointing noun, it will cancel the recognition of the wake-up word in the voice information and instead send the voice information with the device-pointing noun to the hub device.

[0098] After receiving the voice information carrying the device-pointing noun sent by the user or the non-hub device, the hub device identifies the target voice recognition device corresponding to the device-pointing noun.

[0099] In implementation, after receiving the voice information sent by the user, the hub device also identifies whether there is a device-pointing noun in it. After the hub device identifies the presence of a device-pointing noun in the voice information sent by the user, or after receiving a voice information with a device-pointing noun sent by a non-hub device, it will confirm the target voice recognition device corresponding to the device-pointing noun based on the stored correspondence between the voice recognition device and the device-pointing noun.

[0100] When the central device recognizes that the target speech recognition device is the central device, it recognizes and processes the speech information carrying the device-pointing noun, and completes the response after recognizing the wake-up word corresponding to itself.

[0101] During implementation, when the central device recognizes that the target voice recognition device is itself, it will directly recognize and process the voice information carrying the noun pointed to by the device, so as to identify whether there is a wake-up word recorded in its own voice recognition library. When it is recognized that there is, the central device executes the corresponding executable operation based on the correspondence between the stored wake-up word and the executable operation, thereby completing the response.

[0102] When the central device identifies that the target voice recognition device is a non-central device, it sends voice information carrying the device-pointing noun to the target voice recognition device, so that the target voice recognition device recognizes and processes the voice information, and enables the target voice recognition device to complete the response after recognizing the wake-up word corresponding to itself.

[0103] In implementation, after the hub device identifies a non-hub device as a target voice device, it will send voice information carrying device-directed nouns to the target voice recognition device. The target voice device recognizes and processes the voice information carrying device-directed nouns to identify whether there is a wake-up word recorded in its own voice recognition library. If it is recognized to exist, the non-hub device marked as the target voice device executes the corresponding executable operation based on the correspondence between the wake-up word and the executable operation stored in itself, thereby completing the response.

[0104] Furthermore, in another embodiment, when the central device receives multiple voice messages carrying device-pointing nouns within a preset recognition time, in order to avoid repeated processing, only the first received voice message carrying device-pointing nouns will be further recognized and processed. The preset recognition time is similar to the waiting time, which is used to reserve time for the propagation of sound and the processing of signals. Among the above-mentioned multiple voice messages carrying device-pointing nouns, there may be both those obtained directly from the user by the central device through its own MIC array and those received from non-central devices, or there may only be those received by the central device from non-central devices.

[0105] Optionally, in another embodiment, the above method may further include the following contents:

[0106] After receiving the speech recognition library update instruction, the central device or the non-central device updates the currently stored speech recognition library based on the speech recognition library update instruction.

[0107] Among them, the speech recognition library update instruction records the correspondence between the new wake-up word and the executable operation.

[0108] In implementation, when a user wants to update the speech recognition library in a speech recognition device to adapt to different application scenarios, he can issue a speech recognition library update instruction by operating the human-computer interaction interface preset on the speech recognition device. After receiving the speech recognition library update instruction, the hub device or non-hub device completes the update of the speech recognition library stored in itself based on the correspondence between the new wake-up word and the executable operation recorded in the speech recognition library update instruction.

[0109] Optionally, in another embodiment, before the above S204, the following contents may also be included:

[0110] The hub device identifies whether there is an active voice recognition device.

[0111] The active speech recognition device refers to the speech recognition device that made the last response within a preset historical time period.

[0112] In implementation, the hub device can record the working status of itself and each non-hub device. For example, when the hub device or a non-hub device is marked as a responding device, the hub device will mark it as active until a preset historical period has passed or another voice recognition device is marked as active. After processing the actual information to be executed, the hub device will identify whether there is an active device at present.

[0113] When the recognition result is yes, the hub device cancels the comparison of the signal strength value and marks the active voice recognition device as a responding device.

[0114] In implementation, when the hub device identifies the existence of an active device, it can directly mark the active device as a responding device without comparing the signal strength value, thereby helping to improve processing efficiency.

[0115] Based on the above method, an embodiment of the present application also discloses a voice recognition device, which can work as a central device or a non-central device based on user selection.

[0116] Reference Figure 3 , the speech recognition equipment includes:

[0117] The voice information receiving module 31 is used to receive the voice wake-up information sent by the user;

[0118] An information recognition module 32, used to recognize the received voice wake-up information and generate a recognition result corresponding to the device itself;

[0119] An information interaction module 33, used for information interaction with other speech recognition devices;

[0120] The actual information to be executed generating module 34 is used to process and obtain the actual information to be executed based on all the recognition results;

[0121] A responding device marking module 35 is used to identify the target recognition result carrying the largest signal strength value, and mark the hub device or non-hub device that generates the target recognition result as a responding device;

[0122] The response execution module 36 is used to complete the response based on the actual information to be executed.

[0123] Optionally, the voice recognition device further includes:

[0124] The storage module is used to store a speech recognition library corresponding to the speech recognition device itself, in which the speech recognition library records the correspondence between the wake-up word and the executable operation corresponding to the speech recognition device.

[0125] The voice information receiving module 31 is also used to receive voice information;

[0126] The information recognition module 32 is also used to identify whether there is a wake-up word in the voice information after receiving the voice information based on the voice recognition library corresponding to each voice recognition device.

[0127] The information recognition module 32 is also used to authenticate the voice information as voice wake-up information when the corresponding wake-up word is recognized in the voice information.

[0128] Optionally, the voice recognition device further includes:

[0129] The effective recognition result marking module is used to compare the signal strength value carried in each recognition result with a preset strength threshold, and mark the recognition result carrying a signal strength value greater than the strength threshold as a effective recognition result.

[0130] The data statistics module is used to count the wake-up words carried by all valid recognition results to identify the wake-up words with the highest frequency.

[0131] The actual to-be-executed information generating module 34 is specifically configured to generate actual to-be-executed information based on the identified wake-up word with the highest occurrence frequency.

[0132] Optionally, the storage module also stores correspondences between different speech recognition devices and device-pointing nouns.

[0133] The information recognition module 32 is also used to cancel the recognition of the wake-up word in the voice information when it is recognized that there is a device-pointing noun in the received voice information; in this case, if the voice recognition module is a non-central device, the information interaction module 33 therein is also used to send the voice information carrying the device-pointing noun to the central device;

[0134] The information recognition module 32 is also used to recognize the target speech recognition device corresponding to the device-pointing noun after receiving the speech information carrying the device-pointing noun;

[0135] When the voice recognition device is a central device, the information recognition module 32 is also used to recognize and process the voice information carrying the device-pointing noun when the target voice recognition device is identified as a central device, and complete the response through the response execution module 36 after recognizing the wake-up word corresponding to itself; or when the target voice recognition device is identified as a non-central device, the voice information carrying the device-pointing noun is sent to the target voice recognition device through the information interaction module 33.

[0136] Optionally, if within a preset recognition time, the voice recognition device receives multiple voice messages carrying device-pointing nouns through the voice information receiving module 31 or the information interaction module 33, the information recognition module 32 only performs further recognition processing on the first received voice message carrying the device-pointing noun.

[0137] Optionally, the speech recognition device further includes a speech recognition library update module, which is used to update a currently stored speech recognition library based on the speech recognition library update instruction after receiving the speech recognition library update instruction.

[0138] Optionally, the voice recognition device further includes an active device marking module for marking active devices; the responding device marking module 35 is further used to mark the identified active voice recognition device as a responding device.

[0139] The embodiment of the present application also discloses an intelligent terminal, which includes a memory and a processor, wherein the memory stores a computer program that can be loaded by the processor and executes the distributed speech recognition device wake-up method as described above.

[0140] The embodiment of the present application also discloses a computer-readable storage medium, which stores a computer program that can be loaded by a processor and execute the distributed speech recognition device wake-up method as described above. The computer-readable storage medium includes, for example: a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and other media that can store program codes.

[0141] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit the protection scope of the application. Obviously, the described embodiments are only some embodiments of the present application, rather than all embodiments. Based on these embodiments, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present application.

Claims

1. A distributed speech recognition device wake-up method, characterized in that: The method is applied to a distributed collaborative device group consisting of at least two speech recognition devices, wherein the speech recognition devices are divided into a central device and at least one non-central device, and the method comprises: The hub device or non-hub device identifies the received voice wake-up information and generates a recognition result corresponding to the device itself, wherein the recognition result carries a signal strength value of the corresponding voice wake-up information; The non-central device sends the recognition result obtained by its own processing to the central device; The hub device processes and obtains the actual information to be executed based on all the recognition results; the hub device identifies the target recognition result carrying the largest signal strength value, and marks the hub device or non-hub device that generates the target recognition result as a responding device; when the responding device is a hub device, the hub device completes the response based on the actual information to be executed; When the responding device is a non-central device, the central device sends the actual information to be executed to the responding device so that it completes the response, and each of the voice recognition devices stores a voice recognition library corresponding to itself, and the voice recognition library records the corresponding relationship between the wake-up word and the executable operation corresponding to the voice recognition device; Before the hub device or the non-hub device recognizes the received voice wake-up information and generates a recognition result corresponding to the device itself, it includes: After receiving the voice information, the hub device or the non-hub device identifies whether there is a wake-up word in the voice information based on the corresponding voice recognition library; When the hub device or the non-hub device recognizes that there is a wake-up word corresponding to itself in the voice information, the voice information is authenticated as voice wake-up information, and the recognition result also carries the wake-up word recognized from the voice wake-up information; The central device processes all the recognition results to obtain actual information to be executed, which specifically includes: The central device compares the signal strength value carried in each recognition result with a preset strength threshold, and marks the recognition result carrying a signal strength value greater than the strength threshold as a valid recognition result; The central device counts the wake-up words carried by all valid recognition results to identify the wake-up word with the highest frequency of occurrence; The central device generates actual information to be executed based on the identified wake-up word with the highest occurrence frequency.

2. The distributed speech recognition device wake-up method according to claim 1, characterized in that: The speech recognition device stores the correspondence between different speech recognition devices and device-pointing nouns; The method further comprises: When the non-central device recognizes that there is a device-directed noun in the received voice information, it cancels the recognition of the wake-up word in the voice information and sends the voice information carrying the device-directed noun to the central device; After receiving the voice information carrying the device-pointing noun sent by the user or the non-hub device, the hub device identifies the target voice recognition device corresponding to the device-pointing noun; When the hub device recognizes that the target speech recognition device is a hub device, the hub device recognizes and processes the speech information carrying the device-pointing noun, and completes the response after recognizing the wake-up word corresponding to itself; When the central device identifies that the target voice recognition device is a non-central device, it sends the voice information carrying the device-pointing noun to the target voice recognition device, so that the target voice recognition device recognizes and processes the voice information, and enables the target voice recognition device to complete the response after recognizing the wake-up word corresponding to itself.

3. The distributed speech recognition device wake-up method according to claim 2, characterized in that: If the central device receives multiple voice messages carrying the noun pointed by the device within a preset recognition time, the central device only performs further recognition processing on the first received voice message carrying the noun pointed by the device.

4. The distributed speech recognition device wake-up method according to claim 1, characterized in that: The method further comprises: After receiving the speech recognition library update instruction, the central device or non-central device updates the currently stored speech recognition library based on the speech recognition library update instruction, and the speech recognition library update instruction records the correspondence between the new wake-up word and the executable operation.

5. The distributed speech recognition device wake-up method according to claim 1, characterized in that: Before the hub device identifies the target identification result carrying the largest signal strength value and marks the hub device or non-hub device generating the target identification result as a responding device, the method further includes: The central device identifies whether there is a voice recognition device currently in an active state, and the voice recognition device in an active state refers to the voice recognition device that last responded within a preset historical time period; When the recognition result is yes, the hub device cancels the comparison of the signal strength value and marks the active voice recognition device as a responding device.

6. A speech recognition device, characterized in that: The speech recognition device works as a hub device or a non-hub device based on user selection, and the speech recognition device comprises: a speech information receiving module (31) for receiving speech wake-up information sent by the user; An information recognition module (32) is used to recognize the received voice wake-up information and generate a recognition result corresponding to the device itself, wherein the recognition result carries a signal strength value of the corresponding voice wake-up information; An information interaction module (33) is used to interact with other speech recognition devices; an actual information to be executed generation module (34) is used to process and obtain actual information to be executed based on all the recognition results; A response device marking module (35), used for identifying the target recognition result carrying the maximum signal strength value, and marking the hub device or non-hub device generating the target recognition result as a response device; A response execution module (36), used for completing the response based on the actual information to be executed, and when the response device is a hub device, the hub device completes the response based on the actual information to be executed; When the responding device is a non-central device, the central device sends the actual information to be executed to the responding device so that it completes the response, and each of the voice recognition devices stores a voice recognition library corresponding to itself, and the voice recognition library records the corresponding relationship between the wake-up word and the executable operation corresponding to the voice recognition device; Before the hub device or the non-hub device recognizes the received voice wake-up information and generates a recognition result corresponding to the device itself, it includes: After receiving the voice information, the hub device or the non-hub device identifies whether there is a wake-up word in the voice information based on the corresponding voice recognition library; When the hub device or the non-hub device recognizes that there is a wake-up word corresponding to itself in the voice information, the voice information is authenticated as voice wake-up information, and the recognition result also carries the wake-up word recognized from the voice wake-up information; The central device processes all the recognition results to obtain actual information to be executed, which specifically includes: The central device compares the signal strength value carried in each recognition result with a preset strength threshold, and marks the recognition result carrying a signal strength value greater than the strength threshold as a valid recognition result; The central device counts the wake-up words carried by all valid recognition results to identify the wake-up word with the highest frequency of occurrence; The central device generates actual information to be executed based on the identified wake-up word with the highest occurrence frequency.

7. An intelligent terminal, characterized in that: The method comprises a memory and a processor, wherein the memory stores a computer program that can be loaded by the processor and executes any one of the methods of claims 1 to 5.

8. A computer-readable storage medium, characterized in that: A computer program is stored which can be loaded by a processor and execute the method according to any one of claims 1 to 5.

Images