Smart home voice control method and system based on full-duplex voice interaction

Abstract

The application provides a smart home voice control method and system based on full-duplex voice interaction, wherein the method comprises the following steps: step 1: acquiring voice information received by a preset first smart home device in real time; step 2: identifying a process recognition result in the voice information, and determining a device mode of the first smart home device; and step 3: if the device mode is full-duplex voice interaction, acquiring appropriate dialogue corpus, and performing instant voice interaction and voice control according to the dialogue corpus and the process recognition result. The smart home voice control method and system based on full-duplex voice interaction can introduce full-duplex technology when a user uses a smart home device, perform instant voice interaction, and improve the continuity of smart voice interaction without repeatedly waking up a smart assistant. In addition, the introduction of appropriate dialogue corpus improves the suitability of smart voice interaction, and greatly improves the user experience.

Description

Technical Field

[0001] This invention relates to the field of voice processing technology, and in particular to a smart home voice control method and system based on full-duplex voice interaction. Background Technology

[0002] Currently, with the development of IoT and voice recognition technologies, smart homes are increasingly being applied in home life. A terminal with mobile voice control software installed receives local voice information and sends it to a server (e.g., a voice command recognition server) via the network for recognition. The server then returns the voice command, which is matched with the locally set command to obtain the correct control command for voice control of smart home devices.

[0003] Application number CN201911295463.4 discloses a method and apparatus for training offline voice firmware in smart homes. The method includes: a terminal device acquiring a training request for offline voice firmware, the training request including a voice command text to be trained; determining a voice training scheme corresponding to the voice command to be trained based on the voice command text; displaying the voice training scheme to the user so that the user can input voice information according to the voice training scheme; after receiving the user's input voice information, generating offline voice firmware corresponding to the voice command text based on the user's input voice information; and sending the offline voice firmware to an offline voice device so that the offline voice device can store the offline voice firmware and recognize the voice control commands issued by the user based on the offline voice firmware. This method can solve the problems of low voice recognition accuracy and fixed and limited voice semantic definitions in existing offline voice devices, thereby enriching the practical functions of offline voice devices and improving the user experience.

[0004] However, in existing technologies, when users interact with devices via voice, they generally use half-duplex communication technology, which is a question-and-answer mode. This requires repeatedly waking up smart home devices, resulting in low interactivity. Furthermore, during voice training, users are required to input voice information according to the voice training plan, which is not user-friendly and leads to a poor user experience.

[0005] In view of this, the present invention provides a smart home voice control method and system based on full-duplex voice interaction to solve the above problems. Summary of the Invention

[0006] One of the objectives of this invention is to provide a smart home voice control method based on full-duplex voice interaction. When users use smart home devices, full-duplex technology is introduced to enable real-time voice interaction without the need to repeatedly wake up the smart assistant, thus improving the continuity of smart voice interaction. At the same time, the introduction of appropriate dialogue data enhances the suitability of smart voice interaction and greatly improves the user experience.

[0007] The smart home voice control method based on full-duplex voice interaction provided in this invention includes:

[0008] Step 1: Acquire voice information received by the first preset smart home device in real time;

[0009] Step 2: Identify the process recognition result in the voice information, and at the same time, determine the device mode of the first smart home device;

[0010] Step 3: If the device mode is full-duplex voice interaction, obtain suitable dialogue data, and perform real-time voice interaction and voice control based on the dialogue data and the process recognition results.

[0011] Preferably, step 2: recognizing the process recognition result in the voice information includes:

[0012] Based on voiceprint recognition technology, the first target voiceprint in the voice information is identified, and the second target voiceprint belonging to different target persons is determined according to the different voiceprint features of the first target voiceprint.

[0013] Extract the speech information corresponding to the second target voiceprint and use it as the first speech sub-information;

[0014] Extract the first instruction semantics corresponding to the preset wake-up semantics from the first voice sub-information and then parse it. At the same time, obtain the issuance time of the first instruction semantics.

[0015] Obtain the list of control device IDs of the target person, and at the same time, obtain the first device ID of the first smart home device;

[0016] Determine whether the first device ID is in the control device ID list. If so, use the first device ID as the second device ID and the first smart home device corresponding to the second device ID as the second smart home device.

[0017] The semantics of the first instruction, the issuance time, and the second smart home device are used as the process identification result.

[0018] Preferably, the first instruction semantics parsed after extracting the first speech sub-information corresponding to the preset wake-up semantics include:

[0019] The first speech sub-information is expanded on a preset time axis, and the first speech sub-information is analyzed for continuity. If the duration of the interruption of the first speech sub-information is greater than or equal to a preset first threshold, the first speech sub-information is split to obtain multiple second speech sub-information.

[0020] Based on a preset semantic parsing template, the second instruction semantics of the second speech sub-information are parsed;

[0021] The semantics of the second instruction are semantically matched with the wake-up semantics. If the match is found, the corresponding second voice sub-information is used as the third voice sub-information.

[0022] The second speech sub-information following the third speech sub-information is traversed sequentially, and the second instruction semantics of the traversed second speech sub-information is obtained each time it is traversed;

[0023] Integrate all the second instruction semantics to obtain the first instruction semantics.

[0024] Preferably, step 3: performing real-time voice interaction and voice control based on the dialogue corpus and the process recognition result, includes:

[0025] Based on the dialogue corpus and process recognition results, voice interaction information and voice control commands are obtained;

[0026] The voice interaction information is played through a preset voice playback device, and at the same time, the first smart home device controlled by the voice control command is controlled by voice according to the voice control command.

[0027] Preferably, the step of obtaining voice interaction information and voice control commands based on the dialogue corpus and process recognition results includes:

[0028] Determine the initiating and responding corpora in the dialogue corpus;

[0029] Get the current time;

[0030] When the current time reaches the recognition time of the process recognition result, the process recognition result that has reached the recognition time is semantically matched with the initiation semantic corresponding to each initiation corpus to obtain the response corpus of the initiation corpus corresponding to the successfully matched initiation semantic, and used as the interaction corpus;

[0031] Based on a preset interactive voice information generation template, voice interactive information is generated according to the interactive corpus;

[0032] Output the voice interaction information;

[0033] Based on the preset instruction semantic-voice control instruction comparison list and the parsing order of the third instruction semantic in the parsing process recognition result, the voice control instructions corresponding to the third instruction semantic are output sequentially.

[0034] Preferably, step 3: obtaining suitable dialogue data includes:

[0035] Obtain historical connection information of connection paths from multiple corpora;

[0036] Based on the historical connection information, the sum of the historical connection evaluation values ​​for each connection is determined as the first suitability index, and then associated with the corresponding corpus.

[0037] Obtain third-party smart home devices from users of the corpus;

[0038] Based on a preset scenario applicability analysis model, a second suitability index is determined according to the first smart home device and the third smart home device, and then associated with the corresponding corpus.

[0039] Obtain the basis for each update and the number of updates for the corpus;

[0040] Based on the update criteria and the number of updates, a third suitability index is determined and associated with the corresponding corpus;

[0041] The first suitability index, the second suitability index, and the third suitability index associated with the corpus are accumulated and calculated to obtain the suitability score. If the suitability score is greater than or equal to a preset second threshold, the pre-selected dialogue data corresponding to the corpus is taken as the suitable dialogue data.

[0042] Preferably, determining the third suitability index based on the update criteria and the number of updates includes:

[0043] Acquire based on value-characteristic rules;

[0044] Based on the aforementioned criteria value characterization rules, the updated criteria are characterized to obtain a criteria value characteristic distribution map;

[0045] Input the value feature distribution map into a preset value analysis description template to obtain the value analysis description information output by the value analysis description template.

[0046] Based on a pre-defined basis value evaluation database, the basis value of the updated basis is determined according to the basis value analysis description information;

[0047] Obtain the preset fourth suitability index corresponding to the number of updates, and at the same time, obtain the correction weight corresponding to the fourth suitability index based on the value;

[0048] The third suitability index is obtained by weighted fusion of the fourth suitability index and the corresponding modified weight.

[0049] Preferably, the smart home voice control method based on full-duplex voice interaction also includes:

[0050] Obtain the devices to be matched from the first smart home devices;

[0051] Based on the preset matching method corresponding to the device to be matched and the matching signal received by the device to be matched, determine whether the matching is successful;

[0052] If so, the device to be matched will be matched accordingly.

[0053] The smart home voice control system based on full-duplex voice interaction provided in this embodiment of the invention includes:

[0054] The acquisition module is used to acquire voice information received by the preset first smart home device in real time;

[0055] The recognition module is used to recognize the process recognition result in the voice information, and at the same time, determine the device mode of the first smart home device;

[0056] The interaction module is used to acquire suitable dialogue data if the device mode is full-duplex voice interaction, and to perform real-time voice interaction and voice control based on the dialogue data and the process recognition result.

[0057] Preferably, the recognition module recognizes the process recognition result in the voice information, and simultaneously determines the device mode of the first smart home device, including:

[0058] The voiceprint recognition unit is used to identify a first target voiceprint in the voice information based on voiceprint recognition technology, and to determine a second target voiceprint belonging to a different target person based on the different voiceprint features of the first target voiceprint.

[0059] The first extraction unit is used to extract the speech information corresponding to the second target voiceprint and use it as the first speech sub-information;

[0060] The second extraction unit is used to extract the first instruction semantics that corresponds to the preset wake-up semantics in the first voice sub-information, and at the same time, obtain the issuance time of the first instruction semantics.

[0061] The ID acquisition unit is used to acquire a list of control device IDs of the target person, and at the same time, acquire the first device ID of the first smart home device;

[0062] The judgment unit is used to determine whether the first device ID is in the control device ID list. If so, the first device ID is used as the second device ID, and the first smart home device corresponding to the second device ID is used as the second smart home device.

[0063] An integration unit is used to use the semantics of the first instruction, the issuance time, and the second smart home device as the process identification result.

[0064] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description, claims, and drawings.

[0065] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0066] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0067] Figure 1 This is a schematic diagram of a smart home voice control method based on full-duplex voice interaction in an embodiment of the present invention;

[0068] Figure 2 This is a schematic diagram of a smart home voice control system based on full-duplex voice interaction in an embodiment of the present invention. Detailed Implementation

[0069] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0070] This invention provides a smart home voice control method based on full-duplex voice interaction, such as... Figure 1 As shown, it includes:

[0071] Step 1: Real-time acquisition of voice information received by a preset first smart home device; the first smart home device is, for example: smart home appliances and smart home operating devices; the voice information is: the voice information of a person received by the first smart home device;

[0072] Step 2: Identify the process recognition result in the voice information, and simultaneously determine the device mode of the first smart home device; the process recognition result is: the recognition information corresponding to the voice information, such as: the recognized semantics and the recognition time; the device mode is: silent mode and full-duplex voice interaction mode; silent mode is: directly controlling the smart home device by voice based on the voice information; full-duplex voice interaction mode is: the system simulates a dialogue between a person and the user based on the voice information, and controls the smart home device by voice.

[0073] Step 3: If the device mode is full-duplex voice interaction, acquire suitable dialogue data, and perform real-time voice interaction and voice control based on the dialogue data and the process recognition results. The dialogue data consists of multiple dialogue statements acquired through big data analysis. Real-time voice interaction and voice control, for example, if a user says, "The air conditioner is too cold," the smart speaker will interact with the user, issuing the voice message: "The air conditioner temperature setting has been increased to 2 degrees Celsius." Simultaneously, based on IoT technology, the air conditioner temperature setting in the user's room will be increased by 2 degrees Celsius.

[0074] The working principle and beneficial effects of the above technical solution are as follows:

[0075] The system identifies the voice information of the first smart home device acquired in real time, obtains the process recognition result, and simultaneously acquires the device mode of the first smart home device. When the device mode is full-duplex voice interaction, it introduces appropriate dialogue data to perform real-time voice interaction and voice control.

[0076] This application introduces full-duplex technology when users use smart home devices, enabling real-time voice interaction without repeatedly waking up the smart assistant, thus improving the continuity of smart voice interaction. At the same time, the introduction of appropriate dialogue data enhances the suitability of smart voice interaction and greatly improves the user experience.

[0077] In one embodiment, step 2: recognizing the process recognition result in the voice information includes:

[0078] Based on voiceprint recognition technology, a first target voiceprint is identified in the speech information. According to the different voiceprint features of the first target voiceprint, second target voiceprints belonging to different target individuals are determined. The first target voiceprint is the sound wave spectrum carrying speech information. Voiceprint recognition technology is existing and can be implemented. The voiceprint feature is the Mel-frequency cepstral coefficient, and the formula for calculating the Mel-frequency cepstral coefficient is as follows:

[0079]

[0080] Where, τ i (n) is the nth Mel-frequency cepstral coefficient of the i-th frame of the speech information, P(m) is the logarithmic power spectrum of the output signal of the i-th frame of the speech information through the m-th preset triangular filter, M is the number of triangular filters, π is 3.141, and cos is the cosine function.

[0081] Extract the voice information corresponding to the second target voiceprint and use it as the first voice sub-information; the first voice sub-information is: the voice information corresponding to each target person;

[0082] The first instruction semantic is extracted from the first voice sub-information and parsed after corresponding to the preset wake-up semantic. At the same time, the issuance time of the first instruction semantic is obtained. The wake-up semantic is: the preset semantic corresponding to waking up the smart home device, which is set by the user. The first instruction semantic is: the semantic in the language information after the target person wakes up the smart home device. The issuance time is: the time when the voice corresponding to the instruction semantic is issued.

[0083] Obtain the list of control device IDs of the target person, and at the same time, obtain the first device ID of the first smart home device; the list of control device IDs consists of the identifier IDs of the smart home devices that the target person can control; the first device ID consists of the unique identifier ID of the first smart home device.

[0084] Determine whether the first device ID is in the control device ID list. If so, use the first device ID as the second device ID and the first smart home device corresponding to the second device ID as the second smart home device. The second smart home device is a smart home device that can be controlled by the target person.

[0085] The semantics of the first instruction, the issuance time, and the second smart home device are used as the process identification result.

[0086] The working principle and beneficial effects of the above technical solution are as follows:

[0087] Based on voiceprint recognition technology, the first target voiceprint in the voice information is identified. According to the different voiceprint characteristics of the target voiceprint, the first voice sub-information belonging to the target person is determined, avoiding control confusion caused by multiple people using voice control. Wake-up semantics are introduced to determine the first instruction semantic corresponding to the wake-up semantic in the first voice sub-information, instead of blindly recognizing all the semantics of the target person, thus reducing system power consumption. A control device ID list is introduced to determine the second smart home devices that the target person has permission to use, improving the rationality of voice control.

[0088] In one embodiment, the first instruction semantics parsed after extracting the first speech sub-information corresponding to the preset wake-up semantics include:

[0089] The first speech sub-information is expanded on a preset timeline, and a continuity analysis is performed on the first speech sub-information. If the duration of the interruption in the first speech sub-information is greater than or equal to a preset first threshold, the first speech sub-information is split to obtain multiple second speech sub-information. The timeline is a recording system that is sequentially linked in time. The continuity analysis is whether the first speech sub-information is continuous. The interruption duration is the duration of the disappearance of speech in the first speech sub-information. The first threshold is preset manually. The second speech sub-information is a continuous speech segment in the first speech sub-information.

[0090] Based on a preset semantic parsing template, the second instruction semantics of the second speech sub-information are parsed; the semantic parsing template is: constraining the parsing behavior to only acquire semantics, and the parsing is implemented based on semantic analysis technology;

[0091] The semantics of the second instruction are semantically matched with the wake-up semantics. If the match is found, the corresponding second voice sub-information is used as the third voice sub-information. The third voice sub-information is: wake-up voice.

[0092] The second voice sub-information following the third voice sub-information is traversed sequentially. During each traversal, the second instruction semantics of the traversed second voice sub-information are obtained. The second voice sub-information following the third voice sub-information is: the voice information corresponding to the issued instruction.

[0093] Integrate all the second instruction semantics to obtain the first instruction semantics.

[0094] The working principle and beneficial effects of the above technical solution are as follows:

[0095] Generally, when people enter a smart home usage scenario, not all voices are command voices. Semantic acquisition of all voices is inefficient. At the same time, continuously emitted voices are more conducive to semantic extraction. Therefore, a first threshold is introduced. If the duration of the interruption between the first voice sub-information is greater than the first threshold, the first voice sub-information is split to obtain the second voice sub-information, which is beneficial for subsequent semantic extraction. The second voice sub-information after determining the third voice information corresponding to the wake-up semantics is used to extract command semantics, which improves the efficiency of command semantic acquisition.

[0096] In one embodiment, step 3: performing real-time voice interaction and voice control based on the dialogue corpus and the process recognition result includes:

[0097] Based on the dialogue data and process recognition results, voice interaction information and voice control commands are obtained; the voice interaction information is the interactive voice of smart home devices; the voice control commands are the control commands issued by the system to smart home devices after receiving the command semantics.

[0098] The voice interaction information is played through a preset voice playback device. Simultaneously, the first smart home device controlled by the voice control command is controlled via voice commands. The preset voice playback device may be, for example, a speaker pre-installed inside the smart home device.

[0099] The working principle and beneficial effects of the above technical solution are as follows:

[0100] When smart home devices interact with voice, they receive semantic input and then use the received data to train an AI model based on the data in a local corpus. The effectiveness of the AI ​​model's response depends on the quality of its training data. While acquiring massive amounts of data based on big data can improve the response experience of smart devices, it requires a lot of system resources for training, resulting in low training efficiency. Furthermore, if the data is not updated for a long time, the system's interaction effect will be poor. Therefore, a solution is needed.

[0101] Based on appropriate dialogue data and process recognition results, voice interaction information and voice control commands are obtained, improving the suitability of the voice interaction information. The voice interaction information is played through a voice playback device, and the first smart home device controlled by the voice control command is controlled by voice according to the voice control command, further enhancing the interaction effect.

[0102] In one embodiment, obtaining voice interaction information and voice control commands based on the dialogue corpus and process recognition results includes:

[0103] The initiating and responding data in the dialogue corpus are determined; the initiating data is the voice information of the questioner in the dialogue; the responding data is the voice information of the responder in the dialogue.

[0104] Get the current time; the current time is the current time recorded locally.

[0105] When the current time reaches the recognition time of the process recognition result, the process recognition result that has reached the recognition time is semantically matched with the initiation semantics corresponding to each initiation corpus to obtain the response corpus of the initiation corpus corresponding to the successfully matched initiation semantics, and is used as the interaction corpus; the interaction corpus is: corpus of instruction semantics suitable for answering the process recognition result;

[0106] Based on a preset interactive voice information generation template, voice interaction information is generated according to the interactive corpus; the interactive voice information generation template is constrained to generate only interactive voice information and not other information.

[0107] The voice interaction information is output; during output, it is output through a pre-set voice playback device.

[0108] Based on a preset instruction semantic-voice control instruction mapping list and the parsing order of the third instruction semantic in the parsing process recognition result, the voice control instructions corresponding to the third instruction semantic are output sequentially. The instruction semantic-voice control instruction mapping list is a database that stores the correspondence between multiple instruction semantics and voice control instructions; the parsing order is: parsing whichever third instruction semantic is first, and then parsing the next third instruction semantic.

[0109] The working principle and beneficial effects of the above technical solution are as follows:

[0110] If the current time is obtained and the recognition time of the process recognition result is reached, the semantic matching of the process recognition result in the dialogue corpus is determined. An interactive voice information generation template is introduced, and voice interaction information is generated based on the interactive corpus, which improves the suitability of voice interaction information acquisition. A command semantic-voice control command comparison list is introduced, and voice control commands are output sequentially according to the parsing order of the third command semantic in the parsing process recognition result, which improves the rationality of voice control command issuance.

[0111] In one embodiment, step 3: obtaining suitable dialogue data includes:

[0112] Obtain historical connection information for multiple corpora; the connection path is the sharing path of the corpus obtained based on big data; the historical connection information is: which platform and when the connection was made;

[0113] Based on the historical connection information, the sum of the historical connection evaluation values ​​for each connection is determined as the first suitability index, and associated with the corresponding corpus; the historical connection evaluation value is the historical connection platform's evaluation of the corpus.

[0114] Obtain third-party smart home devices from users of the corpus; the user is the party sharing the corpus.

[0115] Based on a pre-defined scenario suitability analysis model, a second suitability index is determined according to the first smart home device and the third smart home device, and then associated with the corresponding corpus. The scenario suitability analysis model is as follows: a neural network model is trained using records of multiple manual analyses of the suitability between scenarios based on different scenario factors, until a convergent artificial intelligence model is established. The second suitability index is the scenario suitability value.

[0116] Obtain the update basis and update count for each update of the corpus; the update basis is: the standard used for the update; the update count is: the number of corpora updated.

[0117] Based on the update criteria and the number of updates, a third suitability index is determined and associated with the corresponding corpus; the more valuable the update criteria and the more updates, the higher the corresponding third suitability index.

[0118] The first suitability index, second suitability index, and third suitability index associated with the corpus are cumulatively calculated to obtain a suitability score. If the suitability score is greater than or equal to a preset second threshold, the pre-selected dialogue data corresponding to the corpus is taken as the suitable dialogue data. The pre-selected dialogue data refers to dialogue data in the corpus; the second threshold is preset manually.

[0119] The working principle and beneficial effects of the above technical solution are as follows:

[0120] When obtaining suitable dialogue data, historical connection information, scenario applicability analysis models, and update criteria and update counts are introduced to jointly determine the suitability, which improves the comprehensiveness and accuracy of the suitability determination criteria and makes it more reasonable.

[0121] In one embodiment, determining the third suitability index based on the update criteria and the number of updates includes:

[0122] The acquisition is based on value characteristic rules; the value characteristic rules are: rules for extracting value characteristic values ​​of the basis in advance, which are, for example: the formulation plan of the updated basis, the identity information of the person who formulates the updated basis, the update cycle, and the acquisition method of the updated content, etc.

[0123] Based on the aforementioned criteria value featureization rules, the updated criteria are characterized to obtain a criteria value feature distribution map; the criteria value feature distribution map includes multiple criteria value feature values, which are arranged according to manually preset rules;

[0124] The value feature distribution map is input into a preset value analysis description template to obtain the value analysis description information output by the value analysis description template. The value analysis description template is preset manually, and the value analysis description information is: description information used to describe the value feature distribution map, such as: a vector constructed based on multiple value feature values ​​in the value feature distribution map.

[0125] Based on a preset basis value evaluation library, the basis value of the updated basis is determined according to the basis value analysis description information. The basis value evaluation library is a database that includes multiple description information and the basis value corresponding to the description information. When determining, the vector constructed based on multiple basis value feature values ​​in the basis value feature distribution map is matched with the description vector corresponding to each basis value in the basis value evaluation library to determine the basis value corresponding to the matching description vector.

[0126] Obtain the preset fourth suitability index corresponding to the number of updates, and at the same time, obtain the correction weight corresponding to the fourth suitability index based on the value; the fourth suitability index is a numerical value, the larger the number of updates, the larger the corresponding fourth suitability index; the larger the value based on, the higher the corresponding correction weight, for example: 1.2;

[0127] The third suitability index is obtained by weighted fusion of the fourth suitability index and its corresponding modified weight. During weighted fusion, the fourth suitability index and its corresponding modified weight are multiplied together.

[0128] The working principle and beneficial effects of the above technical solution are as follows:

[0129] The introduction of value-based characteristic rules to construct a value-based characteristic distribution map improves the standardization of its construction. The introduction of a value analysis description template to determine value analysis description information improves the accuracy of its acquisition. The introduction of a value evaluation database to determine the value of updated criteria, and the determination of the correction weight for the fourth suitability index based on the value, along with the assignment of the corresponding correction weight to the fourth suitability index, yields a corrected fourth suitability index. This corrected fourth suitability index is then used as the third suitability index, further improving the accuracy and rationality of the third suitability index acquisition.

[0130] In one embodiment, the smart home voice control method based on full-duplex voice interaction further includes:

[0131] Obtain the devices to be matched from the first smart home devices; the devices to be matched are: smart home devices that are ready to be matched, such as: smart air conditioners and IoT gateways that are ready to be connected to the Internet of Things;

[0132] Based on the preset matching method corresponding to the device to be matched and the matching signal received by the device to be matched, it is determined whether the matching is successful; the matching method is, for example, Bluetooth matching;

[0133] If so, the device to be matched will be matched accordingly.

[0134] The working principle and beneficial effects of the above technical solution are as follows:

[0135] Matching verification is performed based on the preset matching method corresponding to the device to be matched in the first smart home device and the matching signal received by the device to be matched, thereby improving the suitability of the matching.

[0136] This invention provides a smart home voice control system based on full-duplex voice interaction, such as... Figure 2 As shown, it includes:

[0137] Module 1 is used to acquire voice information received by a preset first smart home device in real time;

[0138] The recognition module 2 is used to recognize the process recognition result in the voice information, and at the same time, determine the device mode of the first smart home device;

[0139] The interaction module 3 is used to acquire suitable dialogue data if the device mode is full-duplex voice interaction, and to perform real-time voice interaction and voice control based on the dialogue data and the process recognition result.

[0140] In one embodiment, the recognition module recognizes the process recognition result in the voice information and simultaneously determines the device mode of the first smart home device, including:

[0141] The voiceprint recognition unit is used to identify a first target voiceprint in the voice information based on voiceprint recognition technology, and to determine a second target voiceprint belonging to a different target person based on the different voiceprint features of the first target voiceprint.

[0142] The first extraction unit is used to extract the speech information corresponding to the second target voiceprint and use it as the first speech sub-information;

[0143] The second extraction unit is used to extract the first instruction semantics that corresponds to the preset wake-up semantics in the first voice sub-information, and at the same time, obtain the issuance time of the first instruction semantics.

[0144] The ID acquisition unit is used to acquire a list of control device IDs of the target person, and at the same time, acquire the first device ID of the first smart home device;

[0145] The judgment unit is used to determine whether the first device ID is in the control device ID list. If so, the first device ID is used as the second device ID, and the first smart home device corresponding to the second device ID is used as the second smart home device.

[0146] An integration unit is used to use the semantics of the first instruction, the issuance time, and the second smart home device as the process identification result.

[0147] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A smart home voice control method based on full-duplex voice interaction, characterized in that, The application comprises the following steps: Step 1: Real-time acquisition of preset voice information received by a first smart home device; Step 2: Recognition of process recognition results in the voice information, and determination of a device mode of the first smart home device; Step 3: If the device mode is full-duplex voice interaction, suitable dialogue corpus is acquired, and instant voice interaction and voice control are performed according to the dialogue corpus and the process recognition results. The step 3: acquisition of suitable dialogue corpus comprises the following steps: Acquisition of historical connection information of connection paths of a plurality of corpus databases; According to the historical connection information, a sum of historical connection evaluation values of each connection is determined as a first suitability index, and is associated with the corresponding corpus database; Acquisition of a third smart home device of a user of the corpus database; Based on a preset scene applicability analysis model, a second suitability index is determined according to the first smart home device and the third smart home device, and is associated with the corresponding corpus database; Acquisition of update basis and update number of each update of the corpus database; Based on the update basis and the update number, a third suitability index is determined and associated with the corresponding corpus database; Cumulative calculation of the first suitability index, the second suitability index and the third suitability index associated with the corpus database obtains a suitability degree, and if the suitability degree is greater than or equal to a preset second threshold value, preselected dialogue corpus corresponding to the corpus database is taken as the suitable dialogue corpus. The step 2: recognition of process recognition results in the voice information comprises the following steps: Based on voiceprint recognition technology, a first target voiceprint in the voice information is recognized, and according to different voiceprint features of the first target voiceprint, a second target voiceprint belonging to different target personnel is determined; The second target voiceprint corresponding voice information is extracted and taken as first voice sub-information; The first instruction semantics after the preset wake-up semantics in the first voice sub-information are extracted, and the sending time of the first instruction semantics is acquired. The step 3: if the device mode is full-duplex voice interaction, suitable dialogue corpus is acquired, and instant voice interaction and voice control are performed according to the dialogue corpus and the process recognition results. ​ ​ 2.The smart home voice control method based on full-duplex voice interaction according to claim 1, wherein, ​ ​ ​ ​ Acquire the target personnel's control device ID list, and acquire the first device ID of the first smart home device; Determine whether the first device ID is in the control device ID list, and if so, take the first device ID corresponding to the first device ID as a second device ID, and take the first smart home device corresponding to the second device ID as a second smart home device; Take the first instruction semantics, the sending time, and the second smart home device as the process recognition result. 3.The smart home voice control method based on full-duplex voice interaction of claim 2, wherein, The first instruction semantics corresponding to the preset wake-up semantics in the first voice sub-information are extracted and analyzed, including: The first voice sub-information is unfolded on a preset time axis, and continuity analysis is performed on the first voice sub-information, if the discontinuous interval length of the first voice sub-information is greater than or equal to a preset first threshold, the first voice sub-information is split to obtain a plurality of second voice sub-information; Based on a preset semantic analysis template, the second instruction semantics of the second voice sub-information are analyzed; The second instruction semantics are matched with the wake-up semantics, and if the match is successful, the second voice sub-information corresponding to the second voice sub-information are taken as third voice sub-information; The second voice sub-information after the third voice sub-information are traversed in turn, and the second instruction semantics of the second voice sub-information traversed each time are acquired; All the second instruction semantics are integrated to obtain the first instruction semantics. 4.The smart home voice control method based on full-duplex voice interaction of claim 1, wherein, The step 3: according to the dialogue corpus and the process recognition result, instant voice interaction and voice control, including: According to the dialogue corpus and the process recognition result, voice interaction information and voice control instructions are acquired; The voice interaction information is played through a preset voice playing device, and the first smart home device controlled by the voice control instructions is controlled according to the voice control instructions. 5.The smart home voice control method based on full-duplex voice interaction according to claim 4, wherein, The voice interaction information and voice control instructions are acquired according to the dialogue corpus and the process recognition result, including: Determine the initiation corpus and the response corpus in the dialogue corpus; Acquire the current time; When the current time reaches the recognition time of the process recognition result, the process recognition result reaching the recognition time is matched with the initiation semantics corresponding to each initiation corpus, the response corpus of the initiation corpus corresponding to the initiation semantics matched successfully is obtained, and is taken as an interactive corpus; Based on a preset interactive voice information generation template, voice interaction information is generated according to the interactive corpus; The voice interaction information is outputted; According to a preset instruction semantics-voice control instruction comparison list and the analysis order of the third instruction semantics in the process recognition result, the voice control instructions corresponding to the third instruction semantics are outputted in turn. 6.The smart home voice control method based on full-duplex voice interaction according to claim 1, wherein, Further comprising: Acquire the first smart home device to be matched; Determine whether the matching is successful according to the preset matching mode corresponding to the to-be-matched device and the matching signal received by the to-be-matched device; If so, the to-be-matched device is matched accordingly.

7. A smart home voice control system based on full duplex voice interaction, characterized in that, Comprising: An acquisition module is configured to acquire voice information received by a preset first smart home device in real time; The recognition module is configured to recognize a process recognition result in the voice information, and determine a device mode of the first smart home device; The interaction module is configured to, if the device mode is full-duplex voice interaction, acquire appropriate dialogue corpus, and perform instant voice interaction and voice control according to the dialogue corpus and the process recognition result; The interaction module performs the following operations: acquire historical connection information of a connection path of a plurality of corpus libraries; determine, according to the historical connection information, a first suitability index as a sum of historical connection evaluation values of each connection, and associate the first suitability index with the corresponding corpus library; acquire a third smart home device of a user of the corpus library; determine, according to the first smart home device and the third smart home device, a second suitability index based on a preset scene suitability analysis model, and associate the second suitability index with the corresponding corpus library; acquire an update basis and an update number of each update of the corpus library; determine, based on the update basis and the update number, a third suitability index, and associate the third suitability index with the corresponding corpus library; cumulatively calculate the first suitability index, the second suitability index and the third suitability index associated with the corpus library to obtain a suitability degree, and if the suitability degree is greater than or equal to a preset second threshold, preselected dialogue corpus corresponding to the corpus library is taken as the appropriate dialogue corpus; The determination of the third suitability index based on the update basis and the update number includes: acquiring a basis value characteristicization rule, wherein the basis value characteristicization rule is a rule for artificially setting a basis value characteristic value in advance, and the basis value characteristic value is a formulation scheme of the update basis, identity information of a person who formulates the update basis, an update cycle and an acquisition path of the update content; characterizing the update basis according to the basis value characteristicization rule to obtain a basis value characteristic distribution diagram; inputting the basis value characteristic distribution diagram into a preset basis value analysis description template to obtain basis value analysis description information output by the basis value analysis description template; determining, based on a preset basis value evaluation library, a basis value of the update basis according to the basis value analysis description information; acquiring a preset fourth suitability index corresponding to the update number, and acquiring a correction weight corresponding to the fourth suitability index according to the basis value; weighting and fusing the fourth suitability index and the correction weight corresponding to the fourth suitability index to obtain the third suitability index.

8. The full-duplex voice interaction based smart home voice control system of claim 7, wherein, The recognition module recognizes a process recognition result in the voice information, and determines a device mode of the first smart home device, including: a voiceprint recognition unit configured to recognize a first target voiceprint in the voice information based on voiceprint recognition technology, and determine a second target voiceprint belonging to different target persons according to different voiceprint characteristics of the first target voiceprint; a first extraction unit configured to extract the voice information corresponding to the second target voiceprint as first voice sub-information; a second extraction unit configured to extract a first instruction semantic parsed after a preset wake-up semantic in the first voice sub-information, and acquire an issuing time of the first instruction semantic. An ID acquisition unit is configured to acquire a control device ID list of the target person, and acquire a first device ID of the first smart home device; A judgment unit is configured to judge whether the first device ID is in the control device ID list, and if so, take the first device ID corresponding to the first device ID as a second device ID, and take the first smart home device corresponding to the second device ID as a second smart home device; An integration unit is configured to take the first instruction semantics, the sending time and the second smart home device as the process recognition result.

Images