Riding falling detection early warning method and device, equipment, storage medium

Abstract

The application discloses a cycling falling detection early warning method and device, equipment and a storage medium, and comprises the following steps: S1, multi-source data synchronous acquisition; S2, cycling state judgment and speed drift correction; S3, audio data noise reduction and uploading; S4, sliding window multi-dimensional motion feature extraction and suspected falling preliminary judgment; S5, road condition adaptive threshold dynamic adjustment; S6, motion and audio dual-dimension verification; S7, rider interaction confirmation and multi-channel early warning for help. The application has the advantages that the cycling falling state can be accurately recognized, the influence of device drift can be effectively eliminated, the detection reliability and emergency response efficiency can be improved, and all-round protection is provided for the cycling safety of the rider.

Description

Technical Field

[0001] This invention belongs to the field of motion state detection and safety early warning technology, specifically relating to a method, device, equipment, and storage medium for detecting and warning of cycling falls. Background Technology

[0002] With the increasing popularity of cycling, safety risks during cycling have become more prominent, and accidental injuries caused by falls are frequent. If a rider loses mobility or consciousness after a fall, they cannot call for help independently and are likely to miss the best opportunity for rescue. To address this issue, various cycling fall warning solutions have emerged in existing technologies. One such solution combines a GNSS positioning module with a six-axis inertial measurement unit to collect speed, acceleration, and attitude angle data. By comprehensively judging the fall status using speed and attitude angle thresholds, a distress signal can be sent through multiple channels.

[0003] However, existing technologies still have many shortcomings: First, fall detection relies on a combination of acceleration and attitude angle thresholds. Long-term use of six-axis inertial measurement units (IMUs) can lead to zero-point drift, distorting angle data and causing false positives or false negatives. Furthermore, additional hardware or algorithms are needed to correct this drift, increasing hardware costs and algorithm complexity. Second, relying solely on motion data for fall detection without combining environmental information for dual verification makes it prone to false alarms due to normal vibrations in complex riding scenarios such as bumpy roads, resulting in insufficient detection accuracy. Third, there is no adaptive detection parameter adjustment mechanism designed for different road conditions. Fixed threshold parameters cannot adapt to diverse riding environments, further increasing the false alarm rate. Fourth, audio data is only used as an auxiliary acquisition item without in-depth analysis. The characteristic audio signals generated by the fall impact are not used to verify the motion data judgment results, failing to fully explore the detection value of multi-source data.

[0004] To address the aforementioned issues, there is an urgent need for a cycling fall detection and early warning solution that can accurately identify cycling falls, effectively eliminate the effects of device drift, combine multi-source data for dual verification, and adapt to complex road conditions. This solution would improve detection reliability and emergency response efficiency, providing comprehensive protection for riders' safety. Summary of the Invention

[0005] In order to overcome the shortcomings of the prior art, the present invention provides a method, device, equipment, and storage medium for detecting and warning of cycling falls.

[0006] To solve the above-mentioned technical problems, the present invention adopts the following technical solution: A method for detecting and issuing early warnings of bicycle falls includes the following steps: S1. Multi-source data synchronous acquisition: The six-axis inertial measurement unit collects the three-dimensional linear acceleration data and three-dimensional angular velocity data of the target rider in real time at a preset acquisition frequency. The six-axis inertial measurement unit integrates a three-axis accelerometer and a three-axis gyroscope. At the same time, the audio acquisition module synchronously collects the audio data of the riding environment, and the GNSS positioning module collects the positioning information and speed-related data of the target rider in real time. S2. Riding Status Judgment and Speed ​​Drift Correction: The riding speed of the target rider is calculated using the linear acceleration data collected by the six-axis inertial measurement unit. The actual speed data collected by the GNSS positioning module is used as a benchmark to compensate for the deviation of the calculated riding speed, eliminating the zero-point drift error of the six-axis inertial measurement unit, and obtaining the corrected riding speed. Based on the corrected riding speed, it is determined whether the target rider is in a riding state. If in a riding state, the fall detection process is initiated; if not in a riding state, the fall detection is paused. S3. Audio data noise reduction and uploading: The environmental audio data collected by the audio acquisition module is subjected to noise reduction processing to filter out invalid noise signals such as wind noise and vibration noise of cycling equipment, so as to obtain effective audio feature data, and the effective audio feature data is uploaded to the cloud server in real time. S4. Multi-dimensional motion feature extraction and preliminary judgment of suspected fall using a sliding window: A sliding time window is set to extract features from the corrected three-dimensional linear acceleration data and three-dimensional angular velocity data, and to monitor in real time the abrupt change features of linear acceleration magnitude, the peak and trough features of angular velocity, and the data stabilization features. When the linear acceleration magnitude exceeds the preset initial abrupt change threshold, and the peak and trough difference of the corresponding angular velocity data exceeds the preset angular velocity threshold, and the linear acceleration data and angular velocity data tend to stabilize to zero within the subsequent preset stabilization time, and the positioning information collected by the GNSS positioning module does not change, it is judged as a suspected fall state. S5. Adaptive Threshold Dynamic Adjustment for Road Conditions: Real-time monitoring of road bump characteristics during cycling. When a continuous N-second period of bumpy road vibration is detected, the linear acceleration sudden change threshold is automatically adjusted to the mean linear acceleration value within the current monitoring period + K times the standard deviation, where N and K are preset adjustment parameters. S6. Dual-dimensional verification of motion and audio: After receiving the valid audio feature data, the cloud server analyzes the valid audio feature data through a pre-trained AI model to identify whether there are characteristic audio signals generated by falling and impact; when the motion data is initially judged to be a suspected fall state and the cloud AI model identifies the fall characteristic audio signal, the dual verification of the fall state is completed. S7. Rider Interaction Confirmation and Multi-Channel Warning and SOS: A warning cancellation prompt is sent to the target rider via the interaction module, and a preset response timer is started simultaneously. If no manual cancellation instruction is received from the target rider within the preset response time, SOS data is sent to the preset contact person and remote server via the communication module, and the SOS information is broadcast via SMS and intelligent robot notification channels. The SOS data includes the target rider's real-time location information, fall detection time, device status data, and collected raw motion data and raw audio data. If the manual cancellation instruction is received, the normal fall detection process is resumed.

[0007] Preferably, in S1, the three-dimensional linear acceleration data collected by the six-axis inertial measurement unit are the linear acceleration values ​​in the X, Y, and Z axes, and the three-dimensional angular velocity data are the angular velocity values ​​in the X, Y, and Z axes; the GNSS positioning module adopts dual-mode positioning of Beidou and GPS.

[0008] Preferably, in S3, the noise reduction process employs an adaptive noise cancellation algorithm, which collects the reference vibration noise of the cycling device as a reference signal and performs cancellation processing with the environmental audio data to filter out invalid noise signals.

[0009] Preferably, in S4, the window duration of the sliding time window is 1 second.

[0010] Preferably, in S5, the criteria for determining the road surface bump characteristics are: the fluctuation amplitude of the linear acceleration modulus value exceeds the preset fluctuation threshold and the fluctuation frequency is greater than the preset frequency within the continuous monitoring period.

[0011] Preferably, in S6, the cloud-pre-trained AI model is trained using a large amount of audio data from cycling scenarios, such as normal cycling audio, fall impact audio, and road bump audio. The pre-trained AI model achieves accurate recognition of fall feature audio signals by extracting the time domain and frequency domain features of the audio. In S7, the warning cancellation prompt includes one or more combinations of sound prompts, vibration prompts, and light prompts; the communication module includes Bluetooth, 4G / 5G, and NB-IoT communication units; the intelligent robot notification channels include WeChat robot and DingTalk robot; the manual cancellation command is sent through the physical button of the riding device or a mobile APP connected via Bluetooth.

[0012] This invention also discloses a cycling fall detection and early warning device to implement the above-mentioned cycling fall detection and early warning method, wherein the cycling fall detection and early warning device includes: The data acquisition module includes a six-axis inertial measurement unit, an audio acquisition module, and a GNSS positioning module. The six-axis inertial measurement unit is used to acquire three-dimensional linear acceleration data and three-dimensional angular velocity data; the audio acquisition module is used to acquire environmental audio data; and the GNSS positioning module is used to acquire positioning information and velocity-related data. The drift correction and status judgment module is connected to the data acquisition module. It is used to calculate the riding speed based on the linear acceleration data, perform zero-point drift correction based on the actual speed data of the GNSS positioning module, and determine whether to start the fall detection process based on the corrected riding speed. An audio preprocessing module, connected to the data acquisition module, is used to perform noise reduction processing on environmental audio data, obtain effective audio feature data, and upload it to a cloud server. The sliding window motion feature extraction module, connected to the drift correction and state judgment module, is used to set a sliding time window to extract multi-dimensional motion features and combine them with GNSS positioning information to make a preliminary judgment of suspected fall. The road condition adaptive adjustment module is connected to the sliding window motion feature extraction module, which is used to monitor the road surface bump characteristics in real time and dynamically adjust the linear acceleration change threshold according to the bump conditions. The cloud-based dual verification module, including a cloud server and a pre-trained AI model, is used to receive valid audio feature data and identify fall feature audio signals, which are then combined with the initial judgment results of motion data to achieve dual verification of the fall status. The interactive confirmation module is connected to the sliding window motion feature extraction module and the cloud dual verification module respectively, and is used to send a warning cancellation prompt and receive the rider's manual cancellation command. The multi-channel early warning module, connected to the interactive confirmation module, is used to send distress data and distress information through multiple channels when no manual cancellation instruction is received.

[0013] Preferably, the six-axis inertial measurement unit is integrated into a cycling headset, cycling computer, or cycling helmet using a patch design; the audio acquisition module is a built-in microphone integrated with the six-axis inertial measurement unit.

[0014] The present invention also discloses a cycling fall detection and early warning device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the above-described cycling fall detection and early warning method.

[0015] The present invention also discloses a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-described method for detecting and warning of bicycle falls.

[0016] By adopting the above technical solution, the present invention has the following beneficial effects: (1) Eliminate the influence of angle drift and reduce hardware costs: This invention abandons the judgment method that relies on attitude angle data in the prior art. It extracts three major motion features, namely the sudden change of linear acceleration modulus, the peak and trough of angular velocity, and the data returning to zero and stabilizing, and combines them with GNSS positioning information to make the initial judgment of fall. This avoids the judgment error caused by the zero point drift of the six-axis inertial measurement unit from the root. There is no need to configure angle correction hardware such as magnetometers, nor is there a need to design complex angle drift correction algorithms. This greatly reduces hardware costs and algorithm complexity, while improving the stability of detection. (2) Dual-dimensional verification of motion and audio significantly reduces false alarm rate: The present invention designs a dual verification mechanism of motion data and cloud audio AI analysis. Only when the motion data is initially judged as a suspected fall and the cloud identifies the fall characteristic audio signal, will the subsequent warning process be entered. This is different from the existing technology that relies on a single dimension of motion data for judgment. It fully explores the detection value of environmental audio data, effectively distinguishes between normal vibration and actual fall impact under complex road conditions, and reduces the false alarm rate under complex road conditions. (3) Adaptive threshold adjustment for road conditions to improve scene adaptability: The present invention sets up a dynamic threshold adjustment mechanism for road conditions, which can automatically adjust the linear acceleration change threshold according to the real-time monitored road bump characteristics, so that the detection parameters can be adapted to different riding road conditions such as flat and bumpy, which solves the problem of detection failure of fixed threshold under complex road conditions in the prior art, and further improves the detection accuracy and the scene adaptability of the solution. (4) Accurate speed drift correction to ensure the accuracy of riding status judgment: This invention introduces the actual speed data of the GNSS positioning module as a benchmark and performs deviation compensation on the riding speed obtained by the integration of the six-axis inertial measurement unit, which effectively eliminates the speed zero drift error, ensures the accuracy of riding status judgment, avoids invalid detection in non-riding state, and reduces equipment power consumption. (5) Interactive confirmation and multi-channel early warning improve the reliability and efficiency of early warning: The present invention sets up a rider interactive confirmation process before sending out distress information to avoid unnecessary disturbance caused by false alarms. At the same time, it sends distress information through multiple channels such as SMS, 4G / 5G, and smart robots, ensuring the effective transmission of distress information in complex outdoor environments. It solves the problem of information transmission failure caused by the single early warning channel in the existing technology, improves the efficiency of emergency response, and provides a guarantee for timely rescue of riders. (6) High hardware integration, strong practicality and promotion: The technical solution of the present invention can be integrated into conventional cycling equipment such as cycling headphones, helmets, and cycling computers. The hardware modification cost is low, the acquisition frequency and algorithm complexity are adapted to the processor performance of cycling equipment, and the operation is simple. Riders do not need to make complex parameter settings. A single charge can achieve long-term operation. It has good practicality and industrialization promotion prospects. In summary, this invention has the advantages of accurately identifying riding falls, effectively eliminating the effects of equipment drift, improving detection reliability and emergency response efficiency, and providing comprehensive protection for rider safety. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating the cycling detection and early warning method of the present invention. Detailed Implementation

[0018] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0019] The components of the embodiments of the invention described and shown in the accompanying drawings can typically be arranged and designed in a variety of different configurations. Therefore, the following detailed description of the embodiments of the invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention.

[0020] Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] In the description of this invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing the invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0022] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0023] Example 1 In this embodiment, the present invention addresses the technical problems of existing cycling fall warning technologies, such as low detection accuracy, high false alarm rate, weak anti-drift capability, and lack of dual verification of multi-source data. It provides a cycling fall detection and warning method, device, equipment, and computer-readable storage medium. The present invention eliminates the influence of angle drift by combining multi-dimensional motion feature analysis, achieves dual verification by combining motion data with cloud-based audio pre-trained AI model analysis, designs a road condition adaptive threshold adjustment mechanism, and integrates multi-channel warnings and rider interaction confirmation processes. This achieves accurate detection and efficient warning of cycling falls while reducing hardware costs and improving the scenario adaptability of the solution.

[0024] like Figure 1 As shown, in one embodiment of the present invention, the first aspect of the present invention provides a method for detecting and warning of bicycle falls, comprising the following steps: S1. Multi-source synchronous data acquisition: A six-axis inertial measurement unit (IMU) collects real-time linear acceleration and angular velocity data of the target rider during riding at a preset acquisition frequency. The IMU integrates a three-axis accelerometer and a three-axis gyroscope. Simultaneously, an audio acquisition module collects audio data of the riding environment, and a GNSS positioning module collects real-time positioning information and speed-related data of the target rider. Specifically, the six-axis IMU collects real-time X, Y, and angular velocity data of the target rider during riding at a preset acquisition frequency of 50Hz-100Hz. The six-axis inertial measurement unit integrates a three-axis accelerometer and a three-axis gyroscope to collect Z-axis linear acceleration data and X, Y, and Z-axis angular velocity data. Simultaneously, it collects audio data of the riding environment through an audio acquisition module and real-time positioning information and speed-related data of the target rider through a GNSS positioning module. The GNSS positioning module uses BeiDou and GPS dual-mode positioning with a positioning accuracy of ≤5m. Alternatively, the six-axis inertial measurement unit can integrate only a three-axis accelerometer (accuracy ±16g) and a three-axis gyroscope (accuracy ±2000rad / s), without collecting any riding posture angle data. S2. Riding Status Judgment and Speed ​​Drift Correction: The riding speed of the target rider is calculated using the linear acceleration data collected by the six-axis inertial measurement unit. The actual speed data collected by the GNSS positioning module is used as a benchmark to compensate for the deviation of the calculated riding speed, eliminating the zero-point drift error of the six-axis inertial measurement unit, and obtaining the corrected riding speed. Based on the corrected riding speed, it is determined whether the target rider is in a riding state. If in a riding state, the fall detection process is initiated; if not in a riding state, the fall detection is paused. The riding speed is calculated by integrating the collected linear acceleration data and combining it with the wheel radius parameters of the riding device to obtain the actual riding speed. The preset riding status judgment threshold is 0.5 m / s. When the corrected riding speed is greater than 0.5 m / s, it is determined to be in a riding state. S3. Audio data noise reduction and uploading: The environmental audio data collected by the audio acquisition module is subjected to noise reduction processing to filter out invalid noise signals such as wind noise and vibration noise of cycling equipment, so as to obtain effective audio feature data. The effective audio feature data is then uploaded to the cloud server in real time. The noise reduction processing adopts an adaptive noise cancellation algorithm, which uses the reference vibration noise of the cycling equipment as a reference signal to cancel the environmental audio data and filter out invalid noise signals. S4. Multi-dimensional motion feature extraction and preliminary judgment of suspected fall using a sliding window: A sliding time window is set to extract features from the corrected three-dimensional linear acceleration data and three-dimensional angular velocity data, and to monitor in real time the abrupt change features of linear acceleration magnitude, the peak and trough features of angular velocity, and the data stabilization features. When the linear acceleration magnitude exceeds the preset initial abrupt change threshold, and the peak and trough difference of the corresponding angular velocity data exceeds the preset angular velocity threshold, and the linear acceleration data and angular velocity data tend to stabilize to zero within the subsequent preset stabilization time, and the positioning information collected by the GNSS positioning module does not change, it is judged as a suspected fall state. Specifically, the sliding time window duration is 1 second, the initial abrupt change threshold for linear acceleration ranges from 15 m / s to 20 m / s, the peak-to-valley difference threshold for angular velocity ranges from 100 rad / s to 150 rad / s, and the preset stabilization time is 2 seconds. When the absolute value of linear acceleration data is ≤0.5 m / s and the absolute value of angular velocity data is ≤0.5 rad / s within two consecutive sliding time windows, it is determined to be approaching zero and stable. When the change in GNSS positioning information is ≤5 m, it is determined to be that the positioning information has not changed. More specifically, a sliding time window of 1 second is set for the corrected... Feature extraction is performed on the three-dimensional linear acceleration data and three-dimensional angular velocity data to monitor three core motion characteristics in real time: linear acceleration magnitude abrupt change characteristics, angular velocity peak and trough characteristics, and data zeroing and stabilization characteristics. When the linear acceleration magnitude exceeds the preset initial abrupt change threshold, and the corresponding angular velocity data peak-trough difference exceeds the preset angular velocity threshold, and the linear acceleration data and angular velocity data tend to stabilize at zero within the subsequent preset stabilization time, and the positioning information collected by the GNSS positioning module remains unchanged, it is determined to be a suspected fall state. In this step, it is not necessary to collect and analyze the riding posture angle data to avoid judgment errors caused by angle zero-point drift. For example, you can make a judgment by following these steps: S4.1 Calculate the linear acceleration magnitude. When the magnitude is >18 m / s², it is determined that the linear acceleration magnitude abrupt change characteristic is met. S4.2 Extract the peak and trough values ​​of angular velocity data. When the peak-trough difference is >120 rad / s, it is determined that the angular velocity peak-trough characteristics are met. S4.3 After satisfying the above two characteristics, monitor the motion data within the next 2 seconds. If the absolute value of linear acceleration data is ≤0.5m / s and the absolute value of angular velocity data is ≤0.5rad / s within two consecutive 1s sliding windows, and the change in GNSS positioning information is ≤5m, it is determined that the data zeroing stationary characteristic is satisfied. S4.4 If all three characteristics mentioned above are met, the condition is determined to be a suspected fall and the process proceeds to the next step.

[0025] S5. Adaptive Threshold Dynamic Adjustment for Road Conditions: Real-time monitoring of road surface bump characteristics during cycling. When continuous bumpy road vibration characteristics are detected for N seconds, the linear acceleration sudden change threshold is automatically adjusted to the mean linear acceleration value + K times the standard deviation within the current monitoring period. N and K are preset adjustment parameters. The preset parameter N ranges from 3s to 5s, and the parameter K ranges from 2 to 3. The criteria for determining the road surface bump characteristics are: the fluctuation amplitude of the linear acceleration modulus value within the continuous monitoring period exceeds the preset fluctuation threshold, and the fluctuation frequency is greater than the preset frequency. S6. Dual-dimensional verification of motion and audio: After receiving the valid audio feature data, the cloud server analyzes the valid audio feature data through a pre-trained AI model to identify whether there are characteristic audio signals generated by falling and impact; when the motion data is initially judged to be a suspected fall state and the cloud AI model identifies the fall characteristic audio signal, the dual verification of the fall state is completed. Specifically, the cloud-based pre-trained AI model is trained using a large amount of audio data from various cycling scenarios, including normal cycling audio, fall and impact audio, and road bump audio. The pre-trained AI model extracts the time and frequency domain features of the audio to accurately identify fall-related audio signals with high accuracy. The pre-trained AI model can be a CNN+LSTM fusion AI model. S7. Rider Interaction Confirmation and Multi-Channel Warning and SOS: A warning cancellation prompt is sent to the target rider via the interaction module, simultaneously initiating a preset response timer. If no manual cancellation command is received from the target rider within the preset response time, SOS data is sent to a preset contact person and a remote server via the communication module, and the SOS information is broadcast via SMS and a smart robot notification channel. The SOS data includes the target rider's real-time location information, fall detection time, device status data, and collected raw motion and audio data. If the manual cancellation command is received, the normal fall detection process resumes. Specifically, the preset response time is 5-10 seconds. The warning cancellation prompt includes one or more combinations of sound, vibration, and light prompts. The communication module includes Bluetooth, 4G / 5G, and NB-IoT communication units. The smart robot notification channels include WeChat and DingTalk robots. The manual cancellation command is sent via the physical button on the riding device or a Bluetooth-connected mobile app. More specifically, the multi-channel early warning and distress call function supports the SOS emergency distress call function. This means that if no manual cancellation command is received from the target rider within the preset response time, riders can send a text message or make a phone call to the set emergency contacts. When a rider is alone, they can use this function to send a distress message to riders with headphones of the same brand (or different brands that support communication function). After receiving the distress message, nearby riders can confirm or cancel the rescue by pressing a button. After the rescue is confirmed, the communication is automatically connected to the team, and the location of the person in distress is automatically shared in real time. The rescuer can click on the avatar of the person being rescued on the map to jump to any map navigation or app navigation. When a rider is in a team of two or more people, they can broadcast to their teammates and confirm if a teammate is in an emergency.

[0026] This invention also discloses a cycling fall detection and early warning device to implement the above-mentioned cycling fall detection and early warning method, wherein the cycling fall detection and early warning device includes: The data acquisition module includes a six-axis inertial measurement unit (IMU), an audio acquisition module, and a GNSS positioning module. The IMU is used to acquire three-dimensional linear acceleration data and three-dimensional angular velocity data; the audio acquisition module is used to acquire environmental audio data; and the GNSS positioning module is used to acquire positioning information and velocity-related data. Specifically, the IMU acquires three-dimensional linear acceleration data and three-dimensional angular velocity data, but does not acquire riding posture angle data; the audio acquisition module has a built-in microphone. The drift correction and status judgment module is connected to the data acquisition module. It is used to calculate the riding speed based on the linear acceleration data, perform zero-point drift correction based on the actual speed data of the GNSS positioning module, and determine whether to start the fall detection process based on the corrected riding speed. An audio preprocessing module, connected to the data acquisition module, is used to perform noise reduction processing on environmental audio data, obtain effective audio feature data, and upload it to a cloud server. The sliding window motion feature extraction module, connected to the drift correction and state judgment module, is used to set a sliding time window to extract multi-dimensional motion features and combine them with GNSS positioning information to make a preliminary judgment of suspected fall. The road condition adaptive adjustment module is connected to the sliding window motion feature extraction module, which is used to monitor the road surface bump characteristics in real time and dynamically adjust the linear acceleration change threshold according to the bump conditions. The cloud-based dual verification module, including a cloud server and a pre-trained AI model, is used to receive valid audio feature data and identify fall feature audio signals, which are then combined with the initial judgment results of motion data to achieve dual verification of the fall status. The interactive confirmation module is connected to the sliding window motion feature extraction module and the cloud dual verification module respectively, and is used to send a warning cancellation prompt and receive the rider's manual cancellation command. The multi-channel early warning module, connected to the interactive confirmation module, is used to send distress data and distress information through multiple channels when no manual cancellation instruction is received.

[0027] The six-axis inertial measurement unit is integrated into cycling headphones, cycling computers, or cycling helmets using a patch design; the audio acquisition module is a built-in microphone integrated with the six-axis inertial measurement unit.

[0028] The present invention also discloses a cycling fall detection and early warning device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the above-described cycling fall detection and early warning method.

[0029] The present invention also discloses a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-described method for detecting and warning of bicycle falls.

[0030] The cycling fall detection and early warning method, device, equipment, and computer-readable storage medium of this invention can be widely applied to various intelligent cycling devices such as cycling headphones, helmets, cycling computers, and smart cycling glasses. It features low hardware modification costs, low algorithm complexity, low power consumption, high detection accuracy, low false alarm rate, and strong scenario adaptability, enabling precise detection and efficient early warning of cycling falls. Furthermore, the technical solution of this invention can be integrated with existing cycling big data platforms and emergency rescue platforms to achieve real-time uploading of fall data and rapid dispatch of rescue resources, effectively reducing the risk of delayed rescue after a cycling fall, ensuring rider safety, and possessing good industrial application prospects and social value, making it suitable for large-scale promotion and application.

[0031] This embodiment does not impose any limitation on the shape, material, structure, etc. of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention shall fall within the protection scope of the technical solution of the present invention.

Claims

1. A method for detecting and issuing early warnings of bicycle falls, characterized in that, Includes the following steps: S1. Multi-source data synchronous acquisition: The six-axis inertial measurement unit collects the three-dimensional linear acceleration data and three-dimensional angular velocity data of the target rider in real time at a preset acquisition frequency. The six-axis inertial measurement unit integrates a three-axis accelerometer and a three-axis gyroscope. At the same time, the audio acquisition module synchronously collects the audio data of the riding environment, and the GNSS positioning module collects the positioning information and speed-related data of the target rider in real time. S2. Riding Status Judgment and Speed ​​Drift Correction: The riding speed of the target rider is calculated using the linear acceleration data collected by the six-axis inertial measurement unit. The actual speed data collected by the GNSS positioning module is used as a reference to compensate for the deviation of the calculated riding speed, eliminate the zero-point drift error of the six-axis inertial measurement unit, and obtain the corrected riding speed. The corrected riding speed is used to determine whether the target rider is riding. If the rider is riding, the fall detection process is initiated. If the rider is not riding, the fall detection is paused. S3. Audio data noise reduction and uploading: The environmental audio data collected by the audio acquisition module is subjected to noise reduction processing to filter out invalid noise signals such as wind noise and vibration noise of cycling equipment, so as to obtain effective audio feature data, and the effective audio feature data is uploaded to the cloud server in real time. S4. Multi-dimensional motion feature extraction and preliminary judgment of suspected fall using a sliding window: A sliding time window is set to extract features from the corrected three-dimensional linear acceleration data and three-dimensional angular velocity data, and to monitor in real time the abrupt change features of linear acceleration magnitude, the peak and trough features of angular velocity, and the data stabilization features. When the linear acceleration magnitude exceeds the preset initial abrupt change threshold, and the peak and trough difference of the corresponding angular velocity data exceeds the preset angular velocity threshold, and the linear acceleration data and angular velocity data tend to stabilize to zero within the subsequent preset stabilization time, and the positioning information collected by the GNSS positioning module does not change, it is judged as a suspected fall state. S5. Adaptive Threshold Dynamic Adjustment for Road Conditions: Real-time monitoring of road bump characteristics during cycling. When a continuous N-second period of bumpy road vibration is detected, the linear acceleration sudden change threshold is automatically adjusted to the mean linear acceleration value within the current monitoring period + K times the standard deviation, where N and K are preset adjustment parameters. S6. Dual-dimensional verification of motion and audio: After receiving the valid audio feature data, the cloud server analyzes the valid audio feature data through a pre-trained AI model to identify whether there are characteristic audio signals generated by falling and impact; when the motion data is initially judged to be a suspected fall state and the cloud AI model identifies the fall characteristic audio signal, the dual verification of the fall state is completed. S7. Rider Interaction Confirmation and Multi-Channel Early Warning and SOS: Send a prompt to the target rider to cancel the early warning through the interaction module, and start the preset response timer at the same time; If no manual release command is received from the target rider within the preset response time, a distress signal is sent to the preset contact person and remote server via the communication module, and the distress information is broadcast via SMS and intelligent robot notification channels. The distress signal includes the target rider's real-time location information, fall detection time, device status data, and collected raw motion data and raw audio data. If the manual release command is received, the normal fall detection process is resumed.

2. The method for detecting and warning of bicycle falls according to claim 1, characterized in that, In S1, the three-axis linear acceleration data collected by the six-axis inertial measurement unit are the linear acceleration values ​​in the X, Y, and Z axes, and the three-axis angular velocity data are the angular velocity values ​​in the X, Y, and Z axes; the GNSS positioning module adopts dual-mode positioning of Beidou and GPS.

3. The method for detecting and warning of bicycle falls according to claim 1, characterized in that, In S3, the noise reduction process adopts an adaptive noise cancellation algorithm, which collects the reference vibration noise of the cycling device as a reference signal and performs cancellation processing with the environmental audio data to filter out invalid noise signals.

4. The method for detecting and warning of bicycle falls according to claim 1, characterized in that, In S4, the window duration of the sliding time window is 1 second.

5. The method for detecting and warning of bicycle falls according to claim 1, characterized in that, In S5, the criteria for determining the road surface bump characteristics are: the fluctuation amplitude of the linear acceleration modulus value exceeds the preset fluctuation threshold and the fluctuation frequency is greater than the preset frequency within the continuous monitoring period.

6. The method for detecting and warning of bicycle falls according to claim 1, characterized in that, In S6, the cloud-pre-trained AI model is trained using a large amount of audio data from cycling scenarios, such as normal cycling audio, fall impact audio, and road bump audio. The pre-trained AI model achieves accurate recognition of fall characteristic audio signals by extracting the time domain and frequency domain features of the audio. In S7, the warning cancellation prompt includes one or more combinations of sound prompts, vibration prompts, and light prompts; the communication module includes Bluetooth, 4G / 5G, and NB-IoT communication units; the intelligent robot notification channels include WeChat robot and DingTalk robot; the manual cancellation command is sent through the physical button of the riding device or a mobile APP connected via Bluetooth.

7. A bicycle fall detection and early warning device, characterized in that, The bicycle fall detection and early warning device, used to implement the bicycle fall detection and early warning method according to claim 1, comprises: The data acquisition module includes a six-axis inertial measurement unit, an audio acquisition module, and a GNSS positioning module. The six-axis inertial measurement unit is used to acquire three-dimensional linear acceleration data and three-dimensional angular velocity data; the audio acquisition module is used to acquire environmental audio data; and the GNSS positioning module is used to acquire positioning information and velocity-related data. The drift correction and status judgment module is connected to the data acquisition module. It is used to calculate the riding speed based on the linear acceleration data, perform zero-point drift correction based on the actual speed data of the GNSS positioning module, and determine whether to start the fall detection process based on the corrected riding speed. An audio preprocessing module, connected to the data acquisition module, is used to perform noise reduction processing on environmental audio data, obtain effective audio feature data, and upload it to a cloud server. The sliding window motion feature extraction module, connected to the drift correction and state judgment module, is used to set a sliding time window to extract multi-dimensional motion features and combine them with GNSS positioning information to make a preliminary judgment of suspected fall. The road condition adaptive adjustment module is connected to the sliding window motion feature extraction module, which is used to monitor the road surface bump characteristics in real time and dynamically adjust the linear acceleration change threshold according to the bump conditions. The cloud-based dual verification module, including a cloud server and a pre-trained AI model, is used to receive valid audio feature data and identify fall feature audio signals, which are then combined with the initial judgment results of motion data to achieve dual verification of the fall status. The interactive confirmation module is connected to the sliding window motion feature extraction module and the cloud dual verification module respectively, and is used to send a warning cancellation prompt and receive the rider's manual cancellation command. The multi-channel early warning module, connected to the interactive confirmation module, is used to send distress data and distress information through multiple channels when no manual cancellation instruction is received.

8. The cycling fall detection and early warning device according to claim 7, characterized in that, The six-axis inertial measurement unit is integrated into cycling headphones, cycling computers, or cycling helmets using a patch design; the audio acquisition module is a built-in microphone integrated with the six-axis inertial measurement unit.

9. A bicycle fall detection and early warning device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the cycling fall detection and early warning method as described in any one of claims 1-6.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the cycling fall detection and early warning method as described in any one of claims 1-6.

Images