CN-121983046-A - Intelligent rescue box object allocation method and system based on voice recognition

CN121983046ACN 121983046 ACN121983046 ACN 121983046ACN-121983046-A

Abstract

The invention discloses a rescue box intelligent object allocation method and a system based on voice recognition, and relates to the technical field of voice recognition, wherein the method comprises the steps of responding to a non-voice trigger signal to automatically enter a ready state, collecting and fusing voice, vision and object state information to perform situation judgment; and enforcing biological characteristics or environment perception verification at key operation nodes, monitoring operation path deviation in real time, adjusting a decision tree path after the operator confirms, and dynamically preparing rescue articles. The invention solves the technical problems of low distribution efficiency and safety caused by low response speed, inaccurate operation guidance and lack of dynamic adaptability of the existing rescue box object distribution method in emergency, achieves the technical effects of realizing quick response, accurate guidance and dynamic adjustment through intelligent voice recognition and multi-mode information fusion, and remarkably improving the distribution efficiency and operation safety of the rescue box object.

Inventors

HE JIAYU
ZHOU YING
ZHU LI
ZHANG DAN
WANG SAN

Assignees

中国人民解放军总医院第六医学中心

Dates

Publication Date: 20260505
Application Date: 20260119

Claims (10)

1. The intelligent rescue box object allocation method based on voice recognition is characterized by comprising the following steps of: Responding to a non-voice environment or an operation trigger signal, so that the rescue box automatically enters a continuous monitoring and ready state without wake-up words; Under the ready state, synchronously acquiring and fusing voice instruction information, visual operation information and article state change information, comprehensively judging the current rescue situation and user intention, and generating situation judgment results; based on the situation judgment result, matching or dynamically generating a decision tree of rescue operation, actively outputting voice guidance information, and simultaneously controlling a rescue box to perform allocation indication and auxiliary execution on a required specific object according to the decision tree; In the executing process of the decision tree, at a preset key operation node, forcibly starting and executing at least one verification process based on biological characteristics or environment perception until verification is passed, and executing the subsequent process; in the rescuing process, the deviation between the actual operation flow and the expected path of the decision tree is monitored in real time, the path of the subsequent decision tree is adjusted in a self-adaptive mode after the operator confirms, and dynamic object use equipment is carried out.
2. The voice recognition-based rescue box intelligent object outfitting method according to claim 1, wherein the non-voice environment or operation trigger signal comprises at least one of an acceleration signal that a rescue box is shifted, a gesture signal that a box door is opened and closed, or a crisis life parameter signal received through communication with an external medical device.
3. The voice recognition-based rescue case intelligent object outfitting method according to claim 1, wherein the visual operation information refers to operator gesture, gesture or line-of-sight direction information acquired through image recognition; the article state change information refers to state information of articles in the box which are taken or put back and are perceived through a weight sensor or a radio frequency identification technology.
4. The voice recognition-based rescue box intelligent object outfitting method of claim 1, wherein comprehensively judging the current rescue situation and the user intention to generate a situation judgment result comprises the following steps: Collecting an environment voice stream in real time through a microphone array, performing voice recognition and semantic understanding, and extracting first key instruction information; Acquiring an operation scene image through a visual sensor, recognizing human body gestures and gestures, and extracting second operation intention information; the method comprises the steps of sensing the taking-out and putting-back states of medicines or instruments in a box in real time through an article state sensor, and obtaining article state information; and carrying out space-time alignment and fusion analysis on the first key instruction information, the second operation intention information and the article state information to generate a situation judgment result of the current rescue stage.
5. The voice recognition-based rescue box smart item provisioning method of claim 1, wherein at least one biometric or context awareness-based verification procedure is forcibly initiated and executed at a preset key operation node, comprising: Presetting at least one forced compliance verification node in an execution path of the decision tree; when the operation flow is perceived to reach any mandatory compliance verification node, stopping the automatic pushing of the decision tree and starting a verification sub-flow; The verification sub-process requires that verification is completed through at least one biological feature or operation evidence, and the verification is allowed to be carried out through subsequent key operation or high-risk article taking through the rear; the verification mode comprises double independent voice verification based on voiceprint recognition or environment safety confirmation based on visual analysis.
6. The voice recognition-based rescue box intelligent service provisioning method of claim 5, wherein the voice recognition-based double independent voice check comprises: When a preset high-risk medicine is required to be taken, the system voice prompt is required to be checked by two persons; Requiring a first operator to recite a check statement and verify a first voiceprint thereof to generate a first voiceprint verification result; In a preset time window, requiring a second operator to recite a check statement, verifying a second voiceprint of the second operator, generating a second voiceprint verification result, and verifying whether the second voiceprint is identical to the first voiceprint; and judging that the verification is passed only when the first voiceprint verification result and the second voiceprint verification result are both passed and the second voiceprint is different from the first voiceprint.
7. The voice recognition-based rescue box smart item provisioning method of claim 5, wherein the visual analysis-based environmental security validation comprises: After the defibrillator is charged, the personnel distribution in the peripheral area of the sickbed is analyzed through the visual sensing module, and whether the personnel return to the safety area is judged; If the person is identified not to fall back to the safety area, continuously sending out a voice warning and preventing discharge execution; And releasing the discharge execution inhibition only when the visual analysis confirms that the personnel positions are all in the safety area or after receiving a command with authority for confirming the voice of the director.
8. The voice recognition-based rescue box intelligent commodity circulation method according to claim 1, wherein real-time monitoring of the expected path deviation of the actual operational flow from the decision tree and adaptive adjustment of the subsequent decision tree path after operator confirmation comprises: in the rescue process, the deviation between the actual operation flow and the decision tree is monitored in real time; When the voice or the operation instruction deviating from the decision tree standard flow is identified, initiating secondary voice confirmation; Dynamically adjusting and updating the current decision tree after the secondary voice confirmation obtains affirmative confirmation; based on the updated decision tree, executing the object allocation of the subsequent decision path; and (3) marking and storing the encryption time stamp of the whole voice stream, the operation event, the verification result and the decision tree change, and generating a structured rescue report according to the encryption time stamp.
9. The voice recognition-based rescue box smart item provisioning method of claim 1, further comprising: In a high-intensity rescue mode, the gain of the microphone array is automatically increased, a directional beam forming algorithm is started, the response priority of non-key voice interaction is reduced, and the command recognition related to rescue is focused.
10. Rescue case intelligence thing is equipped with system based on speech recognition, its characterized in that, the system includes: The situation triggering module is used for responding to a non-voice environment or an operation triggering signal, so that the rescue box automatically enters a continuous monitoring and ready state without wake-up words; the situation judging module is used for synchronously collecting and fusing voice instruction information, visual operation information and article state change information in the ready state, comprehensively judging the current rescue situation and the user intention, and generating a situation judging result; The dynamic decision module is used for matching or dynamically generating a decision tree of rescue operation based on the situation judgment result, actively outputting voice guide information, and controlling a rescue box to perform allocation indication and auxiliary execution on a required specific object according to the decision tree; the forced compliance verification module is used for forcedly starting and executing at least one verification process based on biological characteristics or environment perception at a preset key operation node in the execution process of the decision tree until the verification is passed, and executing the subsequent process; the self-adaptive adjustment module is used for monitoring the expected path deviation between the actual operation flow and the decision tree in real time in the rescue process, and carrying out self-adaptive adjustment on the path of the subsequent decision tree after the operator confirms the actual operation flow and the decision tree so as to carry out dynamic object use equipment.

Description

Intelligent rescue box object allocation method and system based on voice recognition Technical Field The invention relates to the technical field of voice recognition, in particular to a rescue box intelligent object allocation method and system based on voice recognition. Background In modern medical emergency scenes, a series of complex and high-risk operations are often required to be completed in a very short time, and the rescue box supplies closely related to the rescue processes still depend on manual operations, so that the rescue processes are often influenced by factors such as operator judgment, environmental changes, pressure and the like, and operation delay or accuracy reduction are easily caused, particularly in emergency scenes of emergency and high pressure. Although some intelligent devices are used for assisting rescue, most of the intelligent devices rely on single voice control or fixed paths, lack real-time adaptive capacity to environment and operation flow, and cannot flexibly cope with complex rescue requirements. Furthermore, existing systems often lack adequate security verification measures, resulting in potential security hazards for the use of high risk items. Disclosure of Invention The application provides a voice recognition-based intelligent object allocation method and system for a rescue box, which are used for solving the technical problems of low object allocation efficiency and safety caused by low response speed, inaccurate operation guidance and lack of dynamic adaptability of the existing rescue box object allocation method under an emergency condition. The first aspect of the application provides a voice recognition-based intelligent rescue box commodity allocation method, which comprises the steps of responding to a non-voice environment or operation trigger signal, enabling the rescue box to automatically enter a continuous monitoring and ready state without a wake-up word, synchronously collecting and fusing voice instruction information, visual operation information and commodity state change information in the ready state, comprehensively judging the current rescue situation and user intention to generate situation judgment results, matching or dynamically generating a decision tree of the rescue operation based on the situation judgment results, actively outputting voice guide information, controlling the rescue box to allocate and instruct and assist required specific commodities according to the decision tree, forcedly starting and executing at least one verification process based on biological characteristics or environment perception at a preset key operation node in the execution process of the decision tree until the verification is passed, monitoring the deviation of an actual operation flow and an expected path of the decision tree in real time in the rescue process, adaptively adjusting the path of a subsequent decision tree after the operator confirms, and dynamically allocating the commodity. The application provides a second aspect of the intelligent object allocation system of the rescue box based on voice recognition, which comprises a situation triggering module, a situation judging module, a dynamic decision module and a follow-up adjustment module, wherein the situation triggering module is used for responding to a non-voice environment or operation triggering signal to enable the rescue box to automatically enter a continuous monitoring and ready state without awakening words, the situation judging module is used for synchronously collecting and fusing voice instruction information, visual operation information and object state change information in the ready state, comprehensively judging the current rescue situation and user intention to generate situation judging results, the dynamic decision module is used for matching or dynamically generating a decision tree of rescue operation based on the situation judging results and actively outputting voice guiding information, and meanwhile controlling the rescue box to perform allocation indication and auxiliary execution on required specific objects according to the decision tree, the forced compliance verification module is used for forcedly starting and executing at least one verification flow based on biological characteristics or environment perception at a preset key operation node until verification is passed, and the self-adaptive adjustment module is used for monitoring the path deviation of a real-time actual operation flow and the tree and performing self-adaptive allocation on the follow-up path of an expected object during the rescue process. One or more technical schemes provided by the application have at least the following technical effects or advantages: The application provides a method and a system for intelligent object allocation of a rescue box based on voice recognition, which relate to the technical field of voice recognition, enable the rescue box to enter a continuous