CN-121983040-A - Intelligent voice control switch system based on Internet of things and edge calculation

CN121983040ACN 121983040 ACN121983040 ACN 121983040ACN-121983040-A

Abstract

The invention relates to the technical field of Internet of things, edge calculation and intelligent voice control, in particular to an intelligent voice control switch system based on the Internet of things and the edge calculation, which comprises a multi-mode context sensing unit, a voice control switch unit and a voice control switch unit, wherein the multi-mode context sensing unit outputs an original data stream and a plurality of voice instructions to be processed; a real-time state vector generation unit that generates a structured context state vector; a dynamic arbitration logic configuration unit configured to output configured arbitration logic; the invention realizes low-delay local real-time arbitration, avoids the high-delay problem of cloud processing and solves the real-time response problem under the limitation of edge node resources.

Inventors

FU YONGKAI
TIAN XIAOMING
SUN GANG

Assignees

惠州永惠科技有限公司

Dates

Publication Date: 20260505
Application Date: 20251231

Claims (7)

1. An intelligent voice-controlled switching system based on internet of things and edge computing, the system configured to run on an edge computing node and comprising: The multi-mode context sensing unit is configured to receive multiple paths of original voice signals, user voiceprint characteristics, environment sensor data and equipment state data concurrently, and output an original data stream and a plurality of voice instructions to be processed based on the received data; A real-time state vector generation unit configured to receive the raw data stream and generate a structured context state vector; a dynamic arbitration logic configuration unit configured to receive the context state vector and select and activate arbitration logic from a lightweight arbitration policy store based on the vector to output configured arbitration logic; An intended prospective arbitration unit configured to receive the plurality of voice instructions to be processed and the configured arbitration logic, arbitrate the voice instructions to generate a single arbitration decision; A collision-free control sequence generation unit configured to receive the single arbitration decision and convert it into a device control sequence that is sent to the target device; And the arbitration execution feedback unit is configured to monitor the execution state of the device control sequence and generate an execution state feedback signal, and the execution state feedback signal is transmitted back to the multi-mode context sensing unit so as to update the device state data.
2. The intelligent voice-controlled switching system based on internet of things and edge computing of claim 1, wherein the real-time state vector generation unit is configured to generate the structured context state vector, which is configured to: Receiving the original data stream; Applying lightweight processing to the raw data stream to resolve the raw data in real-time to a structured state; Comparing voiceprint features in the original data stream with a pre-stored authorized user list, or comparing noise decibel values in the original data stream with a preset threshold; synthesizing the structured state into a low-dimensional representation with business significance; Generating the context state vector; the dimensions in the context state vector are configured to represent qualitative states or quantitative risks, including high priority user presence, environmental noise, security risk, or high energy consumption conflict states.
3. The intelligent voice-controlled switching system based on internet of things and edge computing of claim 1, wherein the dynamic arbitration logic configuration unit is configured to output the configured arbitration logic configured to: Receiving the context state vector; Selecting and activating matched arbitration logic from a lightweight arbitration policy bank pre-stored in an edge node based on a global snapshot characterized by the context state vector; when the context state vector display environment is highly noisy, judging that the reliability of voiceprint recognition is reduced; triggering arbitration logic which ignores the voice print priority and adopts semantic definition priority instead in response to the reliability reduction; Triggering highest security priority arbitration logic when the context state vector indicates that there is a security risk; the highest security priority arbitration logic is configured to automatically reject all instructions that may exacerbate risk.
4. The intelligent voice-controlled switching system based on internet of things and edge computing of claim 1, wherein the intent-prospective arbitration unit is configured to generate the single arbitration decision by: Applying the configured arbitration logic as a lightweight utility function or cost function; Applying the utility function or cost function to the intent represented by each voice instruction to be processed, including the intent not to be executed; calculating a utility score or an execution cost for the intent to execute in the current context in conjunction with the context state vector; Performing intention look-ahead on each intention, and calculating conflict risks based on preset cost; the lowest cost or highest utility decision is arbitrated and the single arbitration decision is generated.
5. The intelligent voice-controlled switching system based on internet of things and edge computing of claim 2, wherein the real-time state vector generation unit is configured to parse the raw data into a structured state in real time, and is configured to: comparing the voice print characteristics of the user in the original data stream with a pre-stored authorized user list; Comparing the noise decibel value in the original data stream with a preset noise threshold; Comparing the smoke alarm data in the original data stream with a preset smoke threshold value to judge that safety risks exist; checking the state of air conditioning equipment and the state of window equipment in the original data stream to determine whether a high-energy-consumption conflict state exists or not; And respectively assigning the comparison and check results to specific dimensions representing the presence, the environmental noise degree, the safety risk and the high-energy consumption conflict state of the high-priority user in the context state vector.
6. The intelligent voice-controlled switching system based on internet of things and edge computing of claim 3, wherein the dynamic arbitration logic configuration unit is configured to output the configured arbitration logic configured to: The lightweight arbitration policy base is prestored at an edge node, and comprises a plurality of logic modules with low calculation cost aiming at specific scenes, wherein the specific scenes comprise safety, energy saving and user conflict; when the context state vector display environment is highly noisy, judging that the reliability of voiceprint recognition is reduced; Responsive to the reduced confidence level, activating corresponding arbitration logic in the lightweight arbitration policy store for prioritizing semantic clarity; when the context state vector displays the security risk triggered by the smoke alarm, activating corresponding highest security priority arbitration logic in the lightweight arbitration policy bank; The highest security priority arbitration logic is configured to prioritize the execution of security instructions and reject all instructions that may exacerbate the security risk, such as an open ventilator instruction.
7. The intelligent voice-controlled switching system based on internet of things and edge computing of claim 4, wherein the intent-prospective arbitration unit is configured to generate the single arbitration decision by: When an air conditioning opening instruction and a window opening instruction are received, and the context state vector comprises a state that a window is opened; Identifying that the dynamic arbitration logic configuration unit activates energy-saving priority arbitration logic due to the detection of a high-energy-consumption conflict state; Applying a cost function corresponding to the energy-saving priority arbitration logic to perform intention look-ahead calculation; For the air conditioning instruction, searching the opened state of the window, judging the window to be in high-energy-consumption conflict, and calculating high execution cost for the window; for the window opening instruction, searching the opened state of the window, judging the window to be an invalid repeated instruction, and calculating low execution cost for the window opening instruction; Arbitrating a decision to reject the air conditioning instruction and ignore the windowing instruction according to the calculated high execution cost and low execution cost; generating a composite decision including voice feedback that a window is detected to be open, requesting a window to be closed first, as the single arbitration decision.

Description

Intelligent voice control switch system based on Internet of things and edge calculation Technical Field The invention relates to the technical fields of Internet of things, edge calculation and intelligent voice control, in particular to an intelligent voice control switch system based on the Internet of things and the edge calculation. Background In the fields of the Internet of things and intelligent home, intelligent voice control systems deployed on edge computing nodes are increasingly popular, the systems need to process multi-user concurrent voice instructions in real time on edge nodes with limited computing and memory resources and solve potential conflicts among the instructions, traditional solutions or depending on cloud processing introduce unacceptably high delays, or fixed static arbitration rules are adopted locally, the stiff logic cannot be dynamically adjusted according to transient physical environments such as user identities, environmental noise, safety states and the like, so that arbitration robustness is poor, and therefore, how to realize a low-delay and high-adaptability real-time intention arbitration mechanism under the condition of limited edge node resources so as to efficiently solve control problems under concurrent and conflict scenes is a technical problem which is needed to be solved by technicians in the field. Disclosure of Invention The invention aims to provide an intelligent voice control switch system based on the Internet of things and edge calculation, which avoids the problem of poor arbitration robustness caused by high delay caused by the fact that the prior art relies on cloud processing or local rigidifying rules, and can realize efficient and self-adaptive real-time intention arbitration with extremely low delay on edge nodes with limited calculation and memory resources, thereby effectively solving the control problem under the scene of multi-user concurrency and instruction conflict, and specifically adopts the following technical scheme: The multi-mode context sensing unit is configured to receive multiple paths of original voice signals, user voiceprint characteristics, environment sensor data and equipment state data concurrently, and output an original data stream and a plurality of voice instructions to be processed based on the received data; A real-time state vector generation unit configured to receive the raw data stream and generate a structured context state vector; a dynamic arbitration logic configuration unit configured to receive the context state vector and select and activate arbitration logic from a lightweight arbitration policy store based on the vector to output configured arbitration logic; An intended prospective arbitration unit configured to receive the plurality of voice instructions to be processed and the configured arbitration logic, arbitrate the voice instructions to generate a single arbitration decision; A collision-free control sequence generation unit configured to receive the single arbitration decision and convert it into a device control sequence that is sent to the target device; And the arbitration execution feedback unit is configured to monitor the execution state of the device control sequence and generate an execution state feedback signal, and the execution state feedback signal is transmitted back to the multi-mode context sensing unit so as to update the device state data. Optionally, the real-time state vector generating unit is configured to generate the structured context state vector, which is configured to: Receiving the original data stream; Applying lightweight processing to the raw data stream to resolve the raw data in real-time to a structured state; Comparing voiceprint features in the original data stream with a pre-stored authorized user list, or comparing noise decibel values in the original data stream with a preset threshold; synthesizing the structured state into a low-dimensional representation with business significance; Generating the context state vector; the dimensions in the context state vector are configured to represent qualitative states or quantitative risks, including high priority user presence, environmental noise, security risk, or high energy consumption conflict states. Optionally, the dynamic arbitration logic configuration unit is configured to output the configured arbitration logic, and configured to: Receiving the context state vector; Selecting and activating matched arbitration logic from a lightweight arbitration policy bank pre-stored in an edge node based on a global snapshot characterized by the context state vector; when the context state vector display environment is highly noisy, judging that the reliability of voiceprint recognition is reduced; triggering arbitration logic which ignores the voice print priority and adopts semantic definition priority instead in response to the reliability reduction; Triggering highest security priority arbitration logic when th