KNX smart home voice control system based on offline AI

Abstract

The invention discloses a KNX smart home voice control system based on offline AI, which comprises a KNX bus, the KNX bus is connected with a plurality of devices and an embedded server module, the embedded server module is connected with a voice acquisition node module through a wireless local area network, the voice acquisition node module is internally provided with an audio acquisition circuit and a wireless transmission circuit, and the wireless transmission circuit is connected with the KNX bus. Voice information acquired by the audio acquisition circuit is received by the embedded server module through the wireless transmission circuit; the embedded server side module is internally provided with an offline AI voice recognition module, the offline AI voice recognition module receives voice information, converts the voice information into characters and then obtains a user instruction through keyword extraction and comparison, and the user instruction controls corresponding equipment to operate. And meanwhile, an off-line deployment mode is adopted, so that the data and privacy security of the user is fully guaranteed.

Description

Technical Field

[0001] The present invention relates to the technical field of smart home, and more specifically, to a KNX smart home voice control system based on offline AI. Background Art

[0002] With the development of technology, AI technology has shined in more and more industries, and the same is true in the smart home industry. The addition of AI technology will promote the true intelligent development of the smart home industry. The system can identify and judge user behavior through multimodal sensors to facilitate corresponding actions. Currently, in the smart home industry, AI technology is still in its infancy, and typical solutions include smart speaker devices such as Xiaoai Tongxue and Tmall Genie. However, these devices need to transmit user data to the cloud server for centralized calculation and processing. This method is affected by the network to a large extent in terms of device response, and on the other hand, it is worrying about the security of user data. Therefore, the combination of an offline AI model + KNX wired network + wireless network becomes the optimal solution. The KNX wired network ensures the stability and reliability of control signals, the wireless network realizes high-speed data transmission, and the offline AI model realizes intelligent control while ensuring user data security. Summary of the Invention

[0003] In view of the problems in the related art, the present invention proposes a KNX smart home voice control system based on offline AI to overcome the above-mentioned technical problems existing in the existing related technologies.

[0004] To this end, the specific technical solution adopted by the present invention is as follows:

[0005] A KNX smart home voice control system based on offline AI includes: a KNX bus, the KNX bus is connected to several devices and an embedded server module, the embedded server module is connected to a voice acquisition node module through a wireless local area network, the voice acquisition node module is built with an audio acquisition circuit and a wireless transmission circuit, and the voice information collected by the audio acquisition circuit is received by the embedded server module through the wireless transmission circuit; the embedded server module is built with an offline AI voice recognition module, the offline AI voice recognition module receives the voice information, converts the voice information into text, obtains a user instruction through keyword extraction and comparison, and the user instruction controls the corresponding device to operate. The present invention not only includes the characteristics of a traditional KNX smart home system, but also adds a voice programming function, and at the same time adopts an offline deployment method to fully protect the user's data and privacy security.

[0006] Furthermore, the voice acquisition node module includes: a lithium battery, a charging management circuit, an audio acquisition circuit, an MCU microcontroller, an audio playback circuit, a wireless local area network circuit, a key and indicator circuit, and a RAM expansion circuit. The lithium battery is connected to the charging management circuit to provide necessary energy for the voice acquisition node module; the audio acquisition circuit acquires and encodes the received voice information; the MCU microcontroller serves as the control center of the voice acquisition node module and is used to control and manage the voice acquisition node module; the audio playback circuit is used to broadcast user instructions; the wireless local area network circuit realizes data reception and transmission of the voice acquisition node module; the key and indicator circuit is used to provide necessary interaction functions for users; the RAM expansion circuit is used to expand the RAM space of the MCU microcontroller and is used to temporarily store the received voice information. The voice acquisition node module is installed in a non-fixed manner, can be moved freely, and is configured with a charging function.

[0007] Furthermore, the audio acquisition circuit includes a MIC microphone and a CODEC audio encoding chip. The MIC microphone receives voice information and converts the sound signal into an analog electrical signal. The CODEC audio encoding chip receives the electrical signal, converts it into a digital signal, and encodes the digital signal into a specified audio format for transmission to the MCU microcontroller. After receiving the user instruction output by the embedded server module, the MCU microcontroller controls the audio playback circuit to play the user instruction.

[0008] Furthermore, the embedded server module includes: an AIOT processor circuit, a KNX communication circuit, a KNX auxiliary power supply, a power management module, and a wireless local area network circuit. The AIOT processor circuit includes an AIOT processor, and the AIOT processor provides hardware performance support for the offline AI voice recognition module. The KNX communication circuit is used to communicate with devices on the KNX bus. The KNX auxiliary power supply is connected to the power management module to provide power support for the embedded server module. The wireless local area network circuit is used for sending and receiving wireless local area network data.

[0009] Furthermore, the embedded server module includes a wired local area network circuit. The wired local area network circuit is used to connect the KNX smart home voice control system based on offline AI to the Internet to realize network application functions, integrating the KNX wired control method and the wireless transmission method. The wireless transmission method can be wifi, Bluetooth, or zigbee.

[0010] Furthermore, the processing of voice information by the embedded server module includes:

[0011] S1: Extract the voice information through FFT Fourier transform or MFCC Mel-frequency cepstral coefficients to generate the feature vector of the voice information;

[0012] S2: Load the feature vector into a pre-trained offline neural network model and decode it to obtain the text sequence corresponding to the feature vector.

[0013] Further, the processing of the text sequence includes:

[0014] S2.1: Extract keywords from the text sequence;

[0015] S2.2: Perform keyword retrieval based on a pre-established keyword library to determine whether the keyword exists. If it exists, the operation is executed; if not, a failure prompt for execution is sent.

[0016] Further, the operation execution is implemented through the KNX control interface, the KNX programming interface, or the KNX device status reading interface, and the failure prompt for execution is given by the voice acquisition node module to prompt the user to change the instruction.

[0017] The beneficial effects of the present invention are:

[0018] 1. This application has all the characteristics of the traditional KNX smart home system, and on the basis of the traditional KNX smart home system, a voice programming function is added. Users can program and control KNX devices by talking. At the same time, compared with the existing implementation method of terminal + AI cloud service on the market, this application adopts an offline deployment method, which fully guarantees the user's data and privacy security. In the scheme selection, an AIOT processor with a built-in NPU neural network processor and a GPU display processor is used, and a lightweight AI prediction model is selected on the model, which fully ensures the computing performance on the premise of providing basic functions;

[0019] 2. This application realizes the recognition and processing of the user's natural language through the AI model built in the voice acquisition node and the embedded server module, and can control smart home devices without relying on fixed entries;

[0020] 3. The embedded server module of this application enables users to freely program and combine scenarios for KNX devices through voice by using the built-in offline AI model + KNX automation programming module, without the cooperation of professional personnel and a PC;

[0021] 4. The voice acquisition node of this application is connected to the embedded server module through a wireless local area network, so that the voice acquisition node can work at any position covered by the wireless local area network;

[0022] 5. The embedded server module of this application uses an offline AI model, and the collected information is not uploaded to the network. On the premise of ensuring privacy security, AI interaction and intelligent control are realized. Description of the Drawings

[0023] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0024] Figure 1 It is a schematic diagram of module connection of the KNX smart home voice control system based on offline AI according to an embodiment of the present invention;

[0025] Figure 2 It is a schematic diagram of internal module connection of the voice acquisition node module of the KNX smart home voice control system based on offline AI according to an embodiment of the present invention;

[0026] Figure 3 It is a schematic diagram of internal module connection of the embedded server module of the KNX smart home voice control system based on offline AI according to an embodiment of the present invention;

[0027] Figure 4 It is a schematic diagram of the processing process of voice information by the embedded server module of the KNX smart home voice control system based on offline AI according to an embodiment of the present invention;

[0028] Figure 5 It is a schematic diagram of the processing process of the text sequence of the embedded server module of the KNX smart home voice control system based on offline AI according to an embodiment of the present invention. Detailed implementation manners

[0029] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0030] In the implementation of the existing KNX smart home system, professional personnel need to program and configure relevant devices through the ETS software (KNX programming software). All functions of the devices need to be pre-burned into the devices. This results in that if relevant scenario functions need to be modified during the use of these devices, only after-sales service personnel can modify them on-site. This causes huge time and communication costs.

[0031] Moreover, existing KNX smart home systems also have panels similar to voice recognition, but these panels all require pre-set command words, with fixed instructions. Users need to strictly memorize these command words to control related devices, and the panels are non-detachable and can only recognize a fixed distance, causing inconvenience to users.

[0032] Existing KNX smart home systems have limited support for intelligence and can only achieve certain functions through pre-set programming. It is not convenient for control and post-adjustment of special scenarios.

[0033] Currently, there are also solutions in the market to control smart homes by connecting to AI cloud models through smart speakers such as Xiaoai Tongxue and Tmall Genie. However, these solutions all require an internet connection, and the control and processing are all in the cloud, resulting in problems of information and privacy leakage, especially involving sensitive information such as sound and images. At the same time, the implementation of the solutions depends on the docking of third-party smart devices and uses non-standard communication protocols. If the manufacturer goes out of business, it will cause problems for later maintenance.

[0034] To solve the problems of non-intelligence and troublesome construction of KNX smart home systems, and at the same time endow it with functions similar to the AI cloud model, this invention scheme is proposed. While improving the intelligence of the KNX smart home system, it solves the data privacy problem of the AI cloud model. The present invention proposes a KNX smart home voice control system based on offline AI, as Figure 1 shown, including: a KNX bus, the KNX bus is connected to n devices and an embedded server module. The embedded server module is connected to a voice collection node module through a wireless local area network. The KNX bus is reserved with a charging dock for the voice collection node for power replenishment of the voice collection node. The system working process is as follows: The user speaks out their own needs to the voice collection node. The voice collection node collects the user's voice data and sends it to the wireless local area network communication link through the built-in wireless transmission circuit (wifi, Bluetooth). The embedded server wireless receiving circuit receives the corresponding voice information, converts the voice data into text information through the offline ASR (automatic speech recognition) model software installed inside, and then converts the natural language into a control instruction through the offline NLP (natural language processing) model software installed inside the embedded server module. The embedded server module sends this control instruction to the KNX bus, and the corresponding device executes the specified action after receiving the instruction.

[0035] As Figure 2As shown in the figure, the voice acquisition node module includes: a lithium battery, a charging management circuit, an audio acquisition circuit, an MCU microcontroller, an audio playback circuit, a wireless local area network circuit, a key and indicator circuit, and a RAM expansion circuit. The lithium battery is connected to the charging management circuit to provide necessary energy for the voice acquisition node module; the audio acquisition circuit collects and encodes the received voice information; the MCU microcontroller serves as the control center of the voice acquisition node module and is used to control and manage the voice acquisition node module; the audio playback circuit is used to broadcast user instructions; the wireless local area network circuit realizes data reception and transmission of the voice acquisition node module; the key and indicator circuit is used to provide necessary interaction functions for users, such as power on / off and power indication; the RAM expansion circuit is used to expand the RAM space of the MCU microcontroller and is used to temporarily store the received voice information.

[0036] The audio acquisition circuit includes a MIC microphone and a CODEC audio encoding chip. The MIC (microphone) receives the user's voice signal and converts it into an analog electrical signal. After the CODEC (audio encoding chip) collects the analog electrical signal, it converts it into a digital signal and encodes it into a specified audio format for transmission to the MCU (microcontroller). The MCU (microcontroller) then forwards these audio signals to the wireless local area network circuit. After receiving and processing the audio signal, the embedded server module sends an execution result feedback through the wireless local area network. After receiving the result feedback, the MCU (microcontroller) controls the audio playback circuit to play the relevant feedback content.

[0037] As Figure 3 shown in the figure, the embedded server module includes: an AIOT processor circuit, a KNX communication circuit, a KNX auxiliary power supply, a power management module, and a wireless local area network circuit. The AIOT processor circuit includes an AIOT processor. The AIOT processor provides hardware performance support for the offline AI voice recognition module and, at the same time, cooperates with external devices to realize the mutual conversion between KNX and network protocols. The KNX communication circuit is used to communicate with devices on the KNX bus. The KNX auxiliary power supply is connected to the power management module to provide power support for the embedded server module. The wireless local area network circuit is used for sending and receiving wireless local area network data.

[0038] It should be noted that the embedded server module includes a wired local area network circuit, which can be used to connect the KNX smart home voice control system based on offline AI to the Internet to realize network application functions.

[0039] As Figure 4 shown in the figure, the processing of voice information by the embedded server module includes:

[0040] S1: Extract the voice information through FFT Fourier transform or MFCC Mel-frequency cepstral coefficients to generate the feature vector of the voice information;

[0041] S2: Load the feature vector into a pre-trained offline neural network model, start the inference layer interface to perform inference, and finally perform CTC (Connectionist Temporal Classification) decoding on the output acoustic model to obtain the corresponding text sequence.

[0042] As Figure 5 shown, the processing of the text sequence includes:

[0043] S2.1: Extract keywords from the text sequence;

[0044] S2.2: Perform keyword retrieval based on a pre-established keyword library to determine whether the keyword exists. If it exists, the operation is executed; if it does not exist, a failure prompt for execution is sent.

[0045] Specifically, extract keywords from the text sequence. For example, for "Help me turn on the chandelier", the extracted keywords are "turn on" and "chandelier". The device retrieves the keyword library. If the keyword exists, it calls the function interface mapped by the keyword to execute the corresponding operation. The function interfaces include KNX control interface, KNX programming interface, KNX device status reading interface, etc. If the keyword is not retrieved, a failure prompt word is sent through the wireless local area network, and the voice control terminal prompts the user to change the instruction.

[0046] Simple examples of keyword function mapping:

[0047] "Create" "scene": The embedded server module will automatically create an application scene.

[0048] "Chandelier" "join" "Scene 1": The embedded server module will automatically add the chandelier to Scene 1 and download it to the corresponding device.

[0049] "Turn on" "chandelier": The embedded server module will automatically send a turn-on command to the chandelier to turn it on.

[0050] "Report" "temperature": The embedded server will automatically query the temperature sensor device and send the temperature data to the voice control terminal to report the current temperature.

[0051] In summary, this system is equipped with a voice collection node and an embedded server module. It collects audio data through the voice collection node and realizes offline AI functions through the embedded server module. It can perform conversational programming and control on the devices in the KNX network through voice. The AI model used is offline deployed, which fully protects the privacy and security of users. The embedded server module integrates multiple functions such as KNX gateway, KNX programmer, and offline AI inference. In addition to the technical solutions mentioned in the present invention, the system may also use wired network networking methods such as 485 and CAN, and wireless network networking methods such as wifi, Bluetooth, and zigbee.

[0052] It should be noted that the circuits without circuit schematic diagrams shown in this application are all mature technologies, so they will not be elaborated in this application.

[0053] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A KNX smart home voice control system based on offline AI, characterized in that, Including: A KNX bus, the KNX bus connecting several devices and an embedded server module, the embedded server module connecting to a voice collection node module via a wireless local area network, The voice collection node module is built-in with an audio collection circuit and a wireless transmission circuit, and the voice information collected by the audio collection circuit is received by the embedded server module via the wireless transmission circuit; The embedded server module is built-in with an offline AI voice recognition module, the offline AI voice recognition module receiving the voice information, converting the voice information into text, and obtaining a user instruction through keyword extraction and comparison, and the user instruction controlling the operation of the corresponding device.

2. The KNX smart home voice control system based on offline AI according to claim 1, wherein The voice collection node module includes: a lithium battery, a charging management circuit, an audio collection circuit, an MCU microcontroller, an audio playback circuit, a wireless local area network circuit, a key and indicator circuit, and a RAM expansion circuit. The lithium battery is connected to the charging management circuit to provide necessary energy for the voice collection node module; the audio collection circuit collects and encodes the received voice information; the MCU microcontroller serves as the control center of the voice collection node module for controlling and managing the voice collection node module; the audio playback circuit is used to broadcast the user instruction; the wireless local area network circuit realizes data reception and transmission of the voice collection node module; the key and indicator circuit is used to provide necessary interaction functions for the user; the RAM expansion circuit is used to expand the RAM space of the MCU microcontroller for temporarily storing the received voice information.

3. An KNX intelligent home voice control system based on offline AI according to claim 1, characterized in that, The audio collection circuit includes a MIC microphone and a CODEC audio encoding chip. The MIC microphone receives the voice information and converts the sound signal into an analog electrical signal. The CODEC audio encoding chip receives the electrical signal, converts it into a digital signal, and encodes the digital signal into a specified audio format for transmission to the MCU microcontroller. After receiving the user instruction output by the embedded server module, the MCU microcontroller controls the audio playback circuit to play the user instruction.

4. A KNX smart home voice control system based on offline AI according to claim 1, characterized in that, The embedded server module includes: an AIOT processor circuit, a KNX communication circuit, a KNX auxiliary power supply, a power management module, and a wireless local area network circuit. The AIOT processor circuit includes an AIOT processor, and the AIOT processor provides hardware performance support for the offline AI voice recognition module. The KNX communication circuit is used for communicating with devices on the KNX bus. The KNX auxiliary power supply is connected to the power management module to provide power support for the embedded server module. The wireless local area network circuit is used for sending and receiving wireless local area network data.

5. An offline AI-based KNX smart home voice control system according to claim 1, characterized in that, The embedded server module includes a wired local area network circuit, and the wired local area network circuit is used to connect the KNX smart home voice control system based on offline AI to the Internet to realize network application functions.

6. The KNX smart home voice control system based on offline AI according to claim 4, characterized in that, The processing of the voice information by the embedded server module includes: S1: Extracting the voice information through FFT Fourier transform or MFCC Mel-frequency cepstral coefficients to generate a feature vector of the voice information; S2: Loading the feature vector into a pre-trained offline neural network model and decoding to obtain a text sequence corresponding to the feature vector.

7. An offline AI-based KNX smart home voice control system according to claim 6, characterized in that, The processing of the text sequence includes: S2.1: Extracting keywords from the text sequence; S2.2: Conduct keyword retrieval based on a pre-established keyword library, determine whether the keyword exists. If it exists, execute the operation; if not, send a prompt indicating that the execution has failed.

8. An offline AI-based KNX smart home voice control system according to claim 7, characterized in that, The operation execution is achieved through the KNX control interface, the KNX programming interface, or the KNX device status reading interface. The prompt indicating that the execution has failed is given by the voice acquisition node module to prompt the user to change the instruction.

Images