Car accident event monitoring and early warning system and method based on star flash
By utilizing the StarFlash Communication-based vehicle accident monitoring and early warning system, and employing audio acquisition and classification recognition technologies, the system addresses the problem of blind spots in vehicle accident monitoring during severe weather conditions. This enables low-cost, easily deployable vehicle accident early warning systems, thereby improving road traffic safety.
Patent Information
- Application Number
- CN202610229545.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-02-26
- Publication Date
- 2026-05-12
AI Technical Summary
Existing road accident monitoring methods are ineffective in monitoring and warning under adverse weather conditions, leading to chain-reaction accidents, especially in foggy weather when vehicles behind cannot obtain information about accidents ahead in a timely manner.
A vehicle accident monitoring and early warning system based on StarFlash communication is adopted. It collects ambient audio through an audio acquisition device, uses an audio classifier to identify braking sounds and vehicle collision sounds, and combines deceleration warning lights and alarm broadcasts to issue early warnings, thus realizing a low-cost and easy-to-deploy distributed early warning system.
It enables reliable monitoring and early warning of traffic accidents under severe weather conditions, reduces deployment costs, improves the level of proactive protection for road traffic safety, and avoids chain-reaction rear-end collisions.
Smart Images
Figure CN122024448A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of traffic accident early warning technology, specifically relating to a traffic accident event monitoring and early warning system and method based on star flashes. Background Technology
[0002] Existing methods for detecting road traffic accidents include: 1) adding collision sensors to road barriers; 2) installing accelerometers on vehicles; and 3) using video surveillance and image recognition technology. These three methods have the following drawbacks: Traffic accident monitoring based on road fence collision detection is ineffective when vehicles collide without hitting the fence. Acceleration sensors installed on vehicles for accident monitoring are problematic because some vehicles lack response sensors, and accidents don't effectively warn following vehicles. Image recognition technology requires video surveillance along the road, but full coverage is costly and computationally expensive, unsuitable for large-scale deployment, and inadequate for monitoring accidents in foggy conditions. Currently, on highways in foggy weather, following vehicles often fail to brake in time after an accident ahead, a major cause of chain-reaction collisions. Therefore, it is essential to provide a star-flash-based traffic accident monitoring and early warning system and method to address these technical issues. Summary of the Invention
[0003] The technical problem to be solved by the present invention is to provide a star-based vehicle accident monitoring and early warning system and method, which aims to overcome the defects of the existing technology, so as to achieve effective monitoring and early warning even under severe weather conditions, thereby improving the level of early warning.
[0004] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows: A vehicle accident monitoring and early warning system based on StarFlash communication includes a vehicle accident monitoring host, an audio acquisition device, a deceleration warning light, and an alarm broadcast; the vehicle accident monitoring host includes a StarFlash communication module, an alarm linkage module, and a vehicle accident recognition module; the audio acquisition device, deceleration warning light, and alarm broadcast are respectively connected to the vehicle accident monitoring host through the StarFlash communication module; The audio acquisition device is used to collect environmental audio data and send it to the vehicle accident monitoring host via the Star Flash Communication Module; The traffic accident recognition module is used to process and analyze the received audio data to determine whether a traffic accident has occurred. The alarm linkage module is used to control the deceleration warning light to flash and control the alarm broadcast to play a warning voice when the vehicle accident recognition module determines that a vehicle accident has occurred, through the star flash communication module.
[0005] Preferably, the StarScan communication module uses the StarScan SLE protocol to achieve communication connection, supporting low power consumption, low latency, multi-device access, and long-distance communication.
[0006] Preferably, the traffic accident recognition module includes: The audio preprocessing and feature extraction unit is used to perform frame segmentation, filtering, windowing, normalization, and feature extraction on audio data. An audio classifier is used to determine whether an audio sound is a braking sound or a vehicle collision sound based on extracted audio features. The car accident event judgment unit is used to determine whether a car accident event has occurred based on the output of the audio classifier and preset logic.
[0007] Preferably, the audio classifier is a classifier trained based on the K-nearest neighbor algorithm, and its input is the audio feature vector after dimensionality reduction.
[0008] Preferably, the audio preprocessing and feature extraction unit performs the following steps: The audio signal is processed by framing, with a frame length of 150-300ms and an overlap of 50%-75% between adjacent frames; Perform median filtering and windowing on each frame of audio; Calculate the short-time energy of each audio frame, and trigger further feature extraction when the short-time energy exceeds a set multiple of the average energy of historical frames; Extract the Mel frequency cepstral coefficients or Mel spectrum features of each audio frame and perform normalization processing.
[0009] Preferably, the traffic accident event determination unit executes at least one of the following determination logics: If a braking sound is detected, and multiple braking sounds are detected consecutively within a set threshold time, it is determined that a car accident may occur. If a vehicle collision sound is detected within a set threshold time after a braking sound is detected, it is determined that a car accident may have occurred. If a vehicle collision sound is detected and its short-term energy exceeds a set threshold, it is determined that a car accident may have occurred.
[0010] Preferably, the traffic accident monitoring host is further configured to report the traffic accident information to the traffic management platform via the network after determining that a traffic accident has occurred; and to receive an alarm extinguishing command from the traffic management platform to control the deceleration warning lights and alarm broadcasts to stop working.
[0011] Preferably, the method for a vehicle accident monitoring and early warning system based on StarFlash communication includes the following steps: The audio acquisition device continuously collects road environment audio and transmits it to the vehicle accident monitoring host via Star Flash Communication. The audio data is segmented into frames, filtered, and its features are extracted to obtain audio features; The trained audio classifier is used to classify audio features and determine whether they contain braking sounds or vehicle collision sounds. If the classification result includes braking sound or vehicle collision sound, then the pre-defined car accident judgment logic is used to determine whether a car accident has occurred. If a traffic accident is confirmed, the speed reduction warning lights will flash via StarFlash Communication, and a warning voice message will be played via alarm broadcast.
[0012] Preferably, the extraction of the audio features includes: Calculate the short-time energy of each frame of audio and trigger Mel-spectral feature extraction when a short-time energy mutation is detected; Each frame of audio is converted into Mel spectral coefficients and then normalized. The Mel spectrum coefficients are reduced to low-dimensional feature vectors using a pre-defined dimensionality reduction model, which are then used as input to the classifier.
[0013] Preferably, the method further includes: After determining that a traffic accident has occurred, the incident information will be reported to the remote traffic management center; Upon receiving the alarm disabling instruction from the traffic management center, stop issuing warnings and restore normal monitoring status.
[0014] The beneficial effects of this invention are as follows: 1. This invention achieves reliable monitoring and early warning of traffic accidents under adverse weather conditions, effectively solving the problem of existing video surveillance solutions failing in low-visibility environments. Existing technologies, based on image recognition, heavily rely on optical visibility, making it difficult to accurately identify traffic accidents in foggy, rainy, or other adverse weather conditions, resulting in blind spots. This invention utilizes roadside audio acquisition devices to continuously collect ambient sounds and employs feature recognition technology for typical accident sound signatures such as braking and collision sounds to achieve traffic accident assessment without relying on optical information. The system uses satellite communication for data transmission, whose long-distance and high-reliability characteristics ensure stable transmission of monitoring signals to the processing host even under adverse weather conditions. Therefore, this invention is particularly suitable for scenarios such as highways and mountain roads that are susceptible to weather conditions and where video surveillance coverage is insufficient or ineffective, achieving all-weather, all-weather traffic accident detection capabilities.
[0015] 2. A low-cost, easily deployable, and scalable distributed early warning system has been constructed, overcoming the drawbacks of traditional solutions such as high deployment costs, large computing power requirements, and difficulty in achieving wide coverage. Existing solutions based on video analytics or high-density sensor networks suffer from high equipment costs, power supply difficulties, complex communication networking, and heavy backend computing power burdens, making large-scale deployment along long-distance roads difficult. This invention employs a lightweight audio analysis algorithm combined with low-power, multi-connectivity StarFlash communication technology, enabling each monitoring node to have local preprocessing and feature extraction capabilities, significantly reducing the reliance on central server computing power and data transmission volume. Simultaneously, StarFlash communication supports one-to-many connections and long-distance transmission, allowing a single host to manage multiple audio acquisition and early warning devices, greatly simplifying system networking complexity and reducing wiring, power supply, and maintenance costs, providing a feasible technical path for achieving continuous monitoring and early warning along the entire road.
[0016] 3. This invention enables real-time, proactive warnings to vehicles behind, significantly improving the level of proactive road traffic safety protection and compensating for the shortcomings of single-vehicle intelligent warning systems. Existing solutions relying on the vehicle's own sensors can only passively record an accident after it occurs, failing to provide warning information to surrounding vehicles, especially those approaching from behind. This can easily lead to chain-reaction rear-end collisions in low visibility conditions. This invention, through a vehicle accident recognition module, immediately controls the flashing of deceleration warning lights and activates a warning broadcast voice prompt via an alarm linkage module after determining a potential accident. This directly and proactively transmits danger information to vehicles behind, giving them valuable deceleration reaction time. This warning mechanism effectively complements the vehicle's own perception, especially when the driver's vision is limited and reaction time is slow. It uses multiple sound and light methods to forcibly attract attention, effectively preventing secondary accidents and improving the overall road safety resilience. Attached Figure Description
[0017] Figure 1 This is a schematic diagram of the system architecture of the present invention; Figure 2 This is a schematic diagram illustrating a specific usage scenario in an embodiment of the present invention; Figure 3 This is a schematic diagram of the method flow in an embodiment of the present invention; Figure 4 This is a schematic diagram of the audio processing algorithm in an embodiment of the present invention. Detailed Implementation
[0018] Example 1: like Figure 1As shown, a vehicle accident monitoring and early warning system based on StarFlash communication includes a vehicle accident monitoring host, an audio acquisition device, a deceleration warning light, and an alarm broadcast; the vehicle accident monitoring host includes a StarFlash communication module, an alarm linkage module, and a vehicle accident recognition module; the audio acquisition device, deceleration warning light, and alarm broadcast are respectively connected to the vehicle accident monitoring host through the StarFlash communication module; The audio acquisition device is used to collect environmental audio data and send it to the vehicle accident monitoring host via the Star Flash Communication Module; The traffic accident recognition module is used to process and analyze the received audio data to determine whether a traffic accident has occurred. The alarm linkage module is used to control the deceleration warning light to flash and control the alarm broadcast to play a warning voice when the vehicle accident recognition module determines that a vehicle accident has occurred, through the star flash communication module.
[0019] Preferably, the StarScan communication module uses the StarScan SLE protocol to achieve communication connection, supporting low power consumption, low latency, multi-device access, and long-distance communication.
[0020] Preferably, the traffic accident recognition module includes: The audio preprocessing and feature extraction unit is used to perform frame segmentation, filtering, windowing, normalization, and feature extraction on audio data. An audio classifier is used to determine whether an audio sound is a braking sound or a vehicle collision sound based on extracted audio features. The car accident event judgment unit is used to determine whether a car accident event has occurred based on the output of the audio classifier and preset logic.
[0021] Preferably, the audio classifier is a classifier trained based on the K-nearest neighbor algorithm, and its input is the audio feature vector after dimensionality reduction.
[0022] Preferably, the audio preprocessing and feature extraction unit performs the following steps: The audio signal is processed by framing, with a frame length of 150-300ms and an overlap of 50%-75% between adjacent frames; Perform median filtering and windowing on each frame of audio; Calculate the short-time energy of each audio frame, and trigger further feature extraction when the short-time energy exceeds a set multiple of the average energy of historical frames; Extract the Mel frequency cepstral coefficients or Mel spectrum features of each audio frame and perform normalization processing.
[0023] Preferably, the traffic accident event determination unit executes at least one of the following determination logics: If a braking sound is detected, and multiple braking sounds are detected consecutively within a set threshold time, it is determined that a car accident may occur. If a vehicle collision sound is detected within a set threshold time after a braking sound is detected, it is determined that a car accident may have occurred. If a vehicle collision sound is detected and its short-term energy exceeds a set threshold, it is determined that a car accident may have occurred.
[0024] Preferably, the traffic accident monitoring host is further configured to report the traffic accident information to the traffic management platform via the network after determining that a traffic accident has occurred; and to receive an alarm extinguishing command from the traffic management platform to control the deceleration warning lights and alarm broadcasts to stop working.
[0025] Preferably, the method for a vehicle accident monitoring and early warning system based on StarFlash communication includes the following steps: The audio acquisition device continuously collects road environment audio and transmits it to the vehicle accident monitoring host via Star Flash Communication. The audio data is segmented into frames, filtered, and its features are extracted to obtain audio features; The trained audio classifier is used to classify audio features and determine whether they contain braking sounds or vehicle collision sounds. If the classification result includes braking sound or vehicle collision sound, then the pre-defined car accident judgment logic is used to determine whether a car accident has occurred. If a traffic accident is confirmed, the speed reduction warning lights will flash via StarFlash Communication, and a warning voice message will be played via alarm broadcast.
[0026] Preferably, the extraction of the audio features includes: Calculate the short-time energy of each frame of audio and trigger Mel-spectral feature extraction when a short-time energy mutation is detected; Each frame of audio is converted into Mel spectral coefficients and then normalized. The Mel spectrum coefficients are reduced to low-dimensional feature vectors using a pre-defined dimensionality reduction model, which are then used as input to the classifier.
[0027] Preferably, the method further includes: After determining that a traffic accident has occurred, the incident information will be reported to the remote traffic management center; Upon receiving the alarm disabling instruction from the traffic management center, stop issuing warnings and restore normal monitoring status.
[0028] Example 2: This embodiment provides a star-flash-based vehicle accident monitoring and early warning system and method, the technical solution of which is: I. System Composition The traffic accident monitoring and early warning system includes a traffic accident monitoring host, an audio acquisition device, speed reduction warning lights, and an alarm broadcast system. The traffic accident monitoring host includes: StarFlash Communication Module: The host connects to the audio acquisition device, speed reduction warning lights, and alarm broadcast via the StarFlash communication module. The system primarily uses the StarFlash SLE protocol for access, which, compared to Bluetooth, offers advantages such as low power consumption, low latency, high maximum connection count, and long communication distance. Large-scale deployment of vehicle accident monitoring devices on roads may face challenges in two areas: power supply difficulties and large-scale sensor networking. Therefore, using StarFlash communication is the optimal solution for vehicle monitoring devices.
[0029] Alarm linkage module: Based on the output of the vehicle accident recognition module, control the flashing of the deceleration warning light and control the broadcast to play voice alarms.
[0030] The traffic accident recognition module includes an audio classifier, an audio preprocessing and feature extraction module, and a traffic accident event judgment module.
[0031] Audio preprocessing and feature extraction module: performs audio preprocessing and feature extraction on the audio data transmitted by the audio acquisition device.
[0032] Audio classifier: The audio classifier is a trained KNN classifier that judges audio features.
[0033] Traffic accident event detection module: Determines whether a traffic accident has occurred. If it is a traffic accident, it sends a traffic accident event signal to the alarm linkage module.
[0034] Audio acquisition device: includes a microphone and a star-flash communication module.
[0035] Deceleration warning light: Includes warning light and star flash communication module.
[0036] Alarm broadcast: Includes a speaker and a star flash communication module.
[0037] like Figures 2-3 As shown, the specific functions are as follows: Suppose that visibility is low due to heavy fog, and a car accident occurs ahead, but vehicles behind are unable to see the accident due to the fog.
[0038] The audio acquisition device near the lane continuously monitors ambient sounds and sends audio data to the vehicle accident monitoring host via the Star Flash Communication module.
[0039] Car accidents are often accompanied by the sound of sudden braking and a collision.
[0040] The vehicle accident monitoring host continuously receives audio data and analyzes the audio data frame by frame. If there is a short-term energy change in the audio, it is determined that there may be a braking or collision event, and the audio will be further analyzed; otherwise, no action is taken.
[0041] After detecting a short-term energy change in the audio frame, the audio frame is denoised and its features are extracted. The audio features are used as the input to the audio classifier, and the output result is whether the audio frame contains braking sound or vehicle collision sound.
[0042] If the monitoring host detects braking or vehicle collision sounds in the current audio, it enters the "vehicle accident event judgment" stage. If it determines that a vehicle accident may have occurred, the alarm linkage module of the vehicle accident monitoring host will take the following measures: First, a control signal is sent to the deceleration warning light via the star flashing communication module, causing the deceleration warning light to flash rapidly; second, an audio signal is sent to the alarm broadcast via the star flashing communication module, playing a voice prompt indicating that a traffic accident has occurred on the road ahead, requesting vehicles behind to slow down immediately.
[0043] Even better, when the traffic accident monitoring host detects a suspected traffic accident, it can report the accident information via the network, and traffic management personnel will then respond to the incident. After the accident is resolved or it is determined that no accident has occurred, the traffic management personnel send an alarm deactivation command to the traffic accident monitoring host, which then controls the deceleration warning lights to stop flashing and the broadcast to stop transmitting accident warning messages.
[0044] This invention enables timely notification of vehicles behind to slow down when a traffic accident occurs ahead.
[0045] Example 3: like Figure 4 As shown, this embodiment provides the following methods for audio data processing, audio feature extraction, and audio classifier training: The vehicle accident detection unit continuously receives audio data and processes the data as follows: 1. Frame segmentation: The audio is cut into short audio frames of fixed length, 300ms in length. (The default received audio sampling frequency is higher than 20kHz). The time interval between adjacent frames is 150ms (that is, there is a 150ms overlap between two adjacent audio frames).
[0046] 2. Perform median filtering on each frame of audio to suppress audio signal noise.
[0047] 3. Apply a window to each frame of the audio signal using the Hamming window function.
[0048] The preprocessing of audio data includes framing and windowing operations. Framing involves analyzing the audio signal in segments. Based on the actual situation, traffic collision or braking sounds last for more than 1 second, so 150ms is taken as the audio frame length. A Hamming window function is applied to each audio frame to reduce spectral leakage.
[0049] The amplitude of the audio signal is normalized to unify the magnitude and reduce the characteristic differences caused by the distance of the car accident from the audio acquisition device.
[0050] This method extracts temporal features from audio frames using Short Time Energy (STE). The method is as follows: For a sound signal X(m), the short-time energy is: ; Where w(n) is the window function and N is the window length.
[0051] If the short-time energy of the a-th audio frame is much higher than the average value of the previous (a-1) audio frames, for example, if the short-time energy of the a-th audio frame is more than ten times the average value, it is judged that a car accident may have occurred, and the audio is further analyzed.
[0052] Audio feature extraction is performed using the following methods: Mel spectrum calculation: Perform STFT on each frame of audio, and square the modulus of the result to obtain the power spectrum.
[0053] The power spectrum is mapped from Hertz units to Mel scales, and the mapped power spectrum is transformed using a Mel filter bank. The number of Mel filter banks is set to 40, thus obtaining 40 Mel coefficients for each frame of audio.
[0054] Parameter normalization: The obtained audio features are normalized using Z-score.
[0055] Basic preparation for audio feature library: Construction of audio feature library for suspicious car accident events: Audio features are obtained by processing multiple known braking audio and vehicle collision audio, and each audio feature is 40-dimensional data.
[0056] Construction of audio feature library for non-car accident events: Non-vehicle accident audio refers to sounds from vehicle engines and transmission systems, tire friction against the ground, car horns, aerodynamic noise, speed bumps, and vibrations from load-bearing objects. Audio features are obtained by processing non-vehicle accident audio, with each feature being 40-dimensional data.
[0057] The obtained data is dimensionality reduced using the UMAP algorithm. By selecting appropriate parameters, the data points of braking audio, vehicle collision audio and non-accident event audio are clearly classified in three-dimensional space, so that there is a linear function that can distinguish them, and the UMAP model at this time is preserved.
[0058] Training the KNN classifier: The K-nearest neighbor (KNN) algorithm is used to train the 3D data after dimensionality reduction to obtain a KNN classifier for traffic accident event discrimination.
[0059] Audio recognition: The car accident recognition module processes the audio to obtain 40-dimensional audio features.
[0060] The resulting UMAP model was reduced to 3 dimensions.
[0061] The data obtained above is fed into the trained KNN classifier for discrimination. If the classifier determines that the identified audio is likely to be braking or vehicle collision audio, then the current audio is judged to be suspicious car accident audio.
[0062] Judgment of car accident incidents: After the classifier determines that the current audio frame contains audio of a suspected car accident, the car accident recognition module has the following possibilities: If the current audio frame is a braking sound, and the braking sound is continuously detected within a set threshold time, it is determined that there is a prolonged braking sound on the road, and it is determined that a car accident may have occurred.
[0063] If the current audio frame is a braking sound, and a vehicle collision sound is detected within a set threshold time, it is determined that a car accident may have occurred.
[0064] If the current audio frame is a vehicle collision sound and the short-term audio energy is higher than the threshold, it is determined that a car accident may have occurred.
Claims
1. A vehicle accident monitoring and early warning system based on star-flash communication, characterized in that, The system includes a vehicle accident monitoring host, an audio acquisition device, a speed reduction warning light, and an alarm broadcast; the vehicle accident monitoring host includes a star-flash communication module, an alarm linkage module, and a vehicle accident recognition module; the audio acquisition device, speed reduction warning light, and alarm broadcast are each connected to the vehicle accident monitoring host via the star-flash communication module. The audio acquisition device is used to collect environmental audio data and send it to the vehicle accident monitoring host via the Star Flash Communication Module; The traffic accident recognition module is used to process and analyze the received audio data to determine whether a traffic accident has occurred. The alarm linkage module is used to control the deceleration warning light to flash and control the alarm broadcast to play a warning voice when the vehicle accident recognition module determines that a vehicle accident has occurred, through the star flash communication module.
2. The star-flash-based vehicle accident monitoring and early warning system according to claim 1, characterized in that, The StarScan communication module uses the StarScan SLE protocol to achieve communication connection, supporting low power consumption, low latency, multi-device access, and long-distance communication.
3. The star-flash-based vehicle accident monitoring and early warning system according to claim 1, characterized in that, The traffic accident recognition module includes: The audio preprocessing and feature extraction unit is used to perform frame segmentation, filtering, windowing, normalization, and feature extraction on audio data. An audio classifier is used to determine whether an audio sound is a braking sound or a vehicle collision sound based on extracted audio features. The car accident event judgment unit is used to determine whether a car accident event has occurred based on the output of the audio classifier and preset logic.
4. The star-flash-based vehicle accident monitoring and early warning system according to claim 3, characterized in that, The audio classifier is a classifier trained based on the K-nearest neighbor algorithm, and its input is the audio feature vector after dimensionality reduction.
5. A traffic accident monitoring and early warning system based on star flashing as described in claim 3, characterized in that, The audio preprocessing and feature extraction unit performs the following steps: The audio signal is processed by framing, with a frame length of 150-300ms and an overlap of 50%-75% between adjacent frames; Perform median filtering and windowing on each frame of audio; Calculate the short-time energy of each audio frame, and trigger further feature extraction when the short-time energy exceeds a set multiple of the average energy of historical frames; Extract the Mel frequency cepstral coefficients or Mel spectrum features of each audio frame and perform normalization processing.
6. A star-flash-based vehicle accident monitoring and early warning system according to claim 3, characterized in that, The traffic accident event judgment unit executes at least one of the following judgment logics: If a braking sound is detected, and multiple braking sounds are detected consecutively within a set threshold time, it is determined that a car accident may occur. If a vehicle collision sound is detected within a set threshold time after a braking sound is detected, it is determined that a car accident may have occurred. If a vehicle collision sound is detected and its short-term energy exceeds a set threshold, it is determined that a car accident may have occurred.
7. A traffic accident monitoring and early warning system based on star flashing as described in claim 1, characterized in that, The traffic accident monitoring host is also configured to report the traffic accident information to the traffic management platform via the network after determining that a traffic accident has occurred; and to receive alarm extinguishing instructions from the traffic management platform to control the deceleration warning lights and alarm broadcasts to stop working.
8. A method for a vehicle accident monitoring and early warning system based on star-flash communication according to any one of claims 1-7, characterized in that, Includes the following steps: The audio acquisition device continuously collects road environment audio and transmits it to the vehicle accident monitoring host via Star Flash Communication. The audio data is segmented into frames, filtered, and its features are extracted to obtain audio features; The trained audio classifier is used to classify audio features and determine whether they contain braking sounds or vehicle collision sounds. If the classification result includes braking sound or vehicle collision sound, then the pre-defined car accident judgment logic is used to determine whether a car accident has occurred. If a traffic accident is confirmed, the speed reduction warning lights will flash via StarFlash Communication, and a warning voice message will be played via alarm broadcast.
9. The method for a vehicle accident monitoring and early warning system based on star-flash communication according to claim 8, characterized in that, The extraction of the audio features includes: Calculate the short-time energy of each frame of audio and trigger Mel-spectral feature extraction when a short-time energy mutation is detected; Each frame of audio is converted into Mel spectral coefficients and then normalized. The Mel spectrum coefficients are reduced to low-dimensional feature vectors using a pre-defined dimensionality reduction model, which are then used as input to the classifier.
10. The method for a vehicle accident monitoring and early warning system based on star-flash communication according to claim 8, characterized in that, The method further includes: After determining that a traffic accident has occurred, the incident information will be reported to the remote traffic management center; Upon receiving the alarm disabling instruction from the traffic management center, stop issuing warnings and restore normal monitoring status.