Rescue system after accident

Abstract

The invention relates to the field of intelligent automobile safety, provides a design method of a rescue system after an accident, and aims to solve the challenge of rescue after a vehicle collision accident. According to the system, accurate evaluation of passenger injury after an accident is realized by integrating information of the whole vehicle driving process, collision accident information and driver physiological feature data, and an injury prediction result is converted into an understandable text or voice format and is fed back to a vehicle terminal and a medical emergency center. The medical terminal can formulate a rescue scheme according to the information, and the vehicle terminal provides rescue guidance or safety prompts for passersby and assists in preventing injury and deterioration. Through accurate evaluation and real-time guidance, the accident rescue efficiency is remarkably improved, the injury degree is reduced, and an innovative solution is provided for safety guarantee in the field of intelligent automobiles.

Description

Technical Field

[0001] The present invention mainly relates to the field of intelligent vehicle safety and provides an innovative design method for post-accident rescue systems. Background Art

[0002] Traditional accident rescue has the problem of limited information, mainly relying on the active reports of passers-by, occupants or other traffic participants. Such information is often one-sided and limited, and may deviate from the actual accident situation. For some major accidents, the injured may be unable to report actively or provide effective information due to severe injuries or unstable emotions. In addition, although operations such as calling for help can be automatically carried out according to the collision intensity in the current accident rescue scenario, the judgment of information such as the severity of the accident and the severity of occupant injuries still needs to be manually operated, and it is difficult for medical rescue to match the actual accident working conditions, which will have a greater impact on the rescue response time and the allocation of medical resources. Therefore, a post-accident rescue system that can accurately evaluate the degree of occupant injury and provide targeted rescue guidance for rescuers is needed. Summary of the Invention

[0003] In view of this, the present invention provides an innovative design method for post-accident rescue systems. By integrating information on the entire vehicle driving process, collision accident information, and driver physiological characteristics and other data, accurate assessment of occupant injuries after an accident is achieved, and targeted rescue guidance is provided for medical emergency centers and on-site passers-by.

[0004] In the first aspect, the present invention proposes a system for comprehensively and accurately collecting information on the entire accident process. It mainly consists of two modules:

[0005] One is vehicle collision and occupant status detection based on environmental sound recognition: Sound sensors such as Micro-Electro-Mechanical System (MEMS) microphones or piezoelectric microphones are installed at the main stable structures inside and outside the vehicle (such as crossbeams) to reduce the interference of low-frequency noise. When the vehicle is in the driving state, the in-vehicle and out-of-vehicle sound sensors will be enabled. In the preprocessing stage, a high-pass filter is used to remove low-frequency noise, and Mel Frequency Cepstrum Coefficient (MFCC) is used to extract the features of the sound. Among them, the environmental sound can be recognized through the following four types of methods:

[0006] 1. Frequency domain analysis: Based on the Fourier transform, the time-domain signal is converted to the frequency domain

[0007]

[0008] Find the frequency components that may be related to car accidents or human voices by calculating the Power Spectral Density (PSD).

[0009] P(f) = |X(f)| 2

[0010] This method can detect a sudden increase within a specific frequency range to represent a collision or breaking sound; or by analyzing multiple frequency bands, the type and severity of the accident can be further determined.

[0011] 2. Time-domain analysis: Based on the short-time energy and short-time zero-crossing rate of the signal to distinguish different sound sources. Short-time energy:

[0012] Short-time zero-crossing rate: where sgn(x) is the sign function.

[0013] This method can detect sudden changes in sound. For example, a sharp increase in energy may indicate a collision; a change in the zero-crossing rate may reflect the irregularity of the sound, such as scratching or breaking.

[0014] 3. Machine learning-based methods: Different machine learning principles can classify sound signals in different frequency bands. For example, Support Vector Machines (SVM) find the optimal classification boundary by minimizing the following objective function; Random Forest improves classification accuracy by integrating multiple decision trees; Deep learning (Convolutional Neural Network, Long Short-Term Memory Network, etc.): uses a multi-layer neural network for automatic feature extraction and classification of sound signals. Machine learning methods can handle high-dimensional and complex sound features and are more suitable for multi-class and imbalanced data sets.

[0015] 4. Comprehensive methods: For example, first use frequency-domain and time-domain analysis to roughly classify the collected sound signals to reduce the data dimension, and then use machine learning methods for more refined classification and prediction.

[0016] The second is to judge the severity of occupant injury by fusing multi-information input sources: Combining the information that can be actually obtained within the time window of the collision warning stage (such as vehicle dynamics state, occupant physiological characteristics, restraint system state, etc.), a judgment and evaluation of the dynamic response of the vehicle and the occupant and the severity of the injury during the collision process can be given.

[0017] The occupant injury prediction algorithm can be based on a stacked long short-term memory (LSTM) model. This network takes four types of information at the collision condition level, restraint system usage level, ambient sound level, and occupant physiological state level as input variables, and the occupant injury severity information (such as injury level) as the output variable.

[0018] Due to the active operations of the driver to avoid collision accidents and the decision-making uncertainty of the driver himself, which will continuously affect the vehicle's dynamic information, the injury risk prediction algorithm can calculate and process the collected information at a fixed sampling interval (such as 50 ms) until the final injury prediction result is obtained after the collision accident occurs.

[0019] Based on the occupant injury conditions of different severities, it provides guidance for the subsequent post-accident rescue system (for example, minor injuries can only notify the traffic police for handling or vehicle rescue, while serious injuries require immediate dispatch of medical rescue, and further process and analyze the injury prediction results in combination with the big data model to obtain feasible diagnosis and treatment first aid measures).

[0020] Beneficial effects

[0021] An information collection system for the whole process of vehicle collision accidents is proposed. This system uses sensors inside and outside the vehicle to capture various key information before, during, and after the collision in real time.

[0022] The functions of injury prediction and rescue guidance are realized. The big data model can convert the prediction and analysis results into easy-to-understand text or voice information. These information will be transmitted back to the in-vehicle system and the medical emergency center in real time to assist in the rescue.

[0023] It is beneficial to carry out targeted rescue after the accident. The system constructed based on the patent can quickly and accurately provide information about the accident and the victim's status for rescue personnel and other traffic participants, improving the rescue efficiency and ensuring the timeliness and accuracy of the rescue. Brief Description of the Drawings

[0024] Figure 1 It is the overall technical flowchart of a post-accident rescue system

[0025] Figure 2 It is the technical flowchart of judging the severity of occupant injury by fusing multiple information input sources in the cloud Detailed Implementation Manner

[0026] Step 1: Collect information on the whole process of the collision based on sensors inside and outside the vehicle

[0027] Step 1.1: Before the collision, turn on the in-vehicle and out-of-vehicle sound sensors during vehicle driving; the vehicle computer records the vehicle kinematic information such as vehicle speed, acceleration, and heading angle in real time; the intelligent terminal device monitors the physiological state of the occupants in real time.

[0028] Step 1.2: At the moment of collision, based on the multi-source sensor signals, determine whether a collision has occurred, and mark the data in the pre-crash stage and upload it to the cloud in a timely manner.

[0029] Step 1.3: After the collision, on the one hand, continuously monitor the physiological signs of the occupants through the intelligent terminal; on the other hand, based on the in-vehicle unit, send inquiries to other traffic participants to obtain a more intuitive judgment of the occupant injury status. Comprehensively judge whether the occupants have symptoms such as massive bleeding and loss of consciousness.

[0030] Step 2: The cloud filters, analyzes, and integrates the collision information data

[0031] Step 2.1: Filter the data uploaded from the in-vehicle unit with the pre-crash stage as the core. For example, filter out low-frequency noise from the sound information and identify material crushing, airbag deployment, human voices, etc. during the collision.

[0032] Step 2.2: Use algorithms to analyze and extract features from multi-source information separately.

[0033] Step 2.3: Predict the occupant injury status based on the occupant injury prediction algorithm, and further process and analyze the injury prediction results in combination with the big data model (generate easy-to-understand text or voice results).

[0034] Step 3: The cloud model outputs easy-to-understand text or voice results and transmits them back to the in-vehicle unit and the medical emergency center

[0035] Step 3.1: For the medical side, it can be judged whether to issue an emergency rescue according to the injury index prediction results (for example: HIC, N ij , CTI); when rescue is needed, medical staff can prepare the corresponding medical equipment and drugs in advance.

[0036] Step 3.2: For the in-vehicle unit, it can provide guidance for the occupants to save themselves, wake up unconscious occupants, play corresponding rescue guidance information to passers-by, and warn other traffic participants to prevent chain accidents.

Claims

1. A post-accident rescue system, characterized in that: The system comprises: A multi-source sensor module is used to collect various signal data related to vehicle collision events in real time, including but not limited to acceleration data, velocity data, displacement data, occupant posture data, environmental parameters, and video and audio data inside and outside the vehicle; A damage analysis module, connected to the multi-source sensor module, for analyzing and processing the collected multi-source signals and analyzing the damage status of the occupant through a big data model; A rescue guidance module, connected to the injury analysis module, for generating an easily understandable description of the injury status and feasible rescue medical measures according to the injury analysis results; The interactive module is connected to the rescue guidance module and is used to convey the injury status description and rescue medical measures to medical personnel, passers-by or other rescue entities in text or voice form.

2. The system according to claim 1, characterized in that The multi-source sensor module includes: in-vehicle and out-vehicle sensors and devices for real-time acquisition of vehicle dynamic data (such as speed, acceleration, etc.); Sound sensors used to detect occupant status and collision information; intelligent terminal devices used to collect occupant physiological status.

3. The system according to claim 1, characterized in that The damage analysis module includes: A damage analysis model based on stacked long short-term memory (LSTM) technology is used to analyze sensor signals in real time and determine the type and severity of occupant injuries.

4. The system according to claim 1, characterized in that The rescue guidance module includes: The data analysis unit is used to convert the injury analysis results into an easy-to-understand description of the occupant's physical injury status; the guidance generation unit is used to generate targeted rescue and medical guidance plans in combination with the medical rescue database, including first aid treatment methods and priority treatment area recommendations.

5. The system according to claim 1, characterized in that The interaction module includes: A text output unit is used to display the description of the injury status and rescue medical measures in clear and easy-to-read text or transmit them through a communication network; a speech synthesis unit is used to convert the description of the injury status and rescue medical measures into voice prompts.

6. A collision detection and rescue method based on the system according to any one of claims 1 to 5, characterized in that: The following steps are involved: Use multi-source sensor modules to collect multi-source signals related to vehicle collision in real time; Processing the multi-source signals and analyzing the occupant's injury status using a damage analysis module; Using a rescue guidance module to generate an understandable damage description and rescue guidance plan according to the damage status; The injury description and rescue guidance plan are output to the rescue subject in the form of text or voice using the interactive module.

7. The method according to claim 6, characterized in that The rescue guidance scheme also includes a recommendation on the priority of medical resource allocation based on the ranking of the severity of the occupant's injuries.

Images