Helmet wearing intelligent detection method and device and electronic equipment

Abstract

The invention provides a helmet wearing intelligent detection method and device and electronic equipment, and the method comprises the steps: obtaining a first acceleration signal of a vehicle and a second acceleration signal of a helmet; based on a preset frequency band selection strategy, the same frequency band is selected from the frequency domain signal of the first acceleration signal and the frequency domain signal of the second acceleration signal, a first frequency band signal and a second frequency band signal are obtained, and the first frequency band signal and the second frequency band signal are frequency bands with the human head active motion component energy proportion lower than a preset threshold value; determining characteristic parameters of a neck transfer function based on the first frequency band signal and the second frequency band signal; determining coherence coefficient characteristics of the first acceleration signal and the second acceleration signal based on components of the first acceleration signal and the second acceleration signal in the same frequency band; and determining whether the helmet is in a wearing state or not based on the feature parameters of the neck transfer function, the coherence coefficient features and a pre-trained classifier. According to the invention, the accuracy of helmet wearing detection can be improved.

Description

Technical Field

[0001] This invention belongs to the field of vehicle safe driving technology, specifically relating to a helmet wearing intelligent detection method, device, and electronic equipment. Background Technology

[0002] With the increasing prevalence of two-wheeled vehicles (such as motorcycles and electric bikes) and construction vehicles, riding safety and work safety are receiving growing attention. Helmets, as crucial equipment for head protection, are directly related to personal safety. However, in practice, the phenomenon of not wearing helmets or wearing them improperly persists despite repeated prohibitions, leading to frequent safety accidents.

[0003] In related technologies, the helmet's motion state is determined by an accelerometer built into it. However, when the vehicle is traveling on a bumpy road, the helmet's motion is highly coupled with the vehicle's, making it difficult to distinguish whether the helmet is worn on the head or placed on the vehicle, resulting in a high rate of misjudgment. Summary of the Invention

[0004] In view of this, the purpose of the present invention is to provide a method, device and electronic device for intelligent detection of helmet wearing, so as to meet the need to improve the accuracy of helmet wearing detection.

[0005] To achieve the above objectives, the present invention provides the following technical solution: According to a first aspect, the present invention provides a helmet-wearing intelligent detection method, comprising: acquiring a first acceleration signal of a vehicle and a second acceleration signal of a helmet during vehicle operation; selecting the same frequency band from the frequency domain signals of the first acceleration signal and the second acceleration signal based on a pre-set frequency band selection strategy to obtain a first frequency band signal and a second frequency band signal, wherein the first frequency band signal and the second frequency band signal are frequency bands in which the energy proportion of the active motion component of the human head is lower than a preset threshold; determining the characteristic parameters of the neck transfer function based on the first frequency band signal and the second frequency band signal; determining the coherence coefficient characteristics of the first acceleration signal and the second acceleration signal based on the components of the first acceleration signal and the second acceleration signal in the same frequency band; and determining whether the helmet is being worn based on the characteristic parameters of the neck transfer function, the coherence coefficient characteristics, and a pre-trained classifier.

[0006] Optionally, based on a pre-set frequency band selection strategy, the same frequency band is selected from the frequency domain signals of the first acceleration signal and the second acceleration signal to obtain a first frequency band signal and a second frequency band signal. This includes: obtaining a target frequency band based on the vehicle's current position, wherein the target frequency band is pre-determined based on the road surface characteristics at the current position when the vehicle was on the previous target road segment, and the previous target road segment is the road segment that the vehicle traversed during the target time period before the current position; and selecting the target frequency band from both the frequency domain signals of the first acceleration signal and the second acceleration signal to obtain the first frequency band signal and the second frequency band signal.

[0007] Optionally, based on the components of the first acceleration signal and the second acceleration signal within the same frequency band, the coherence coefficient characteristics of the two signals are determined, including: performing frequency band filtering and inverse transformation on the first acceleration signal and the second acceleration signal to obtain a first frequency band time domain signal and a second frequency band time domain signal; performing frame segmentation processing on the first frequency band time domain signal and the second frequency band time domain signal to obtain multiple first acceleration time frames and multiple second acceleration time frames; transforming the first frequency band time domain signal and the second frequency band time domain signal within each time frame to the frequency domain to obtain a first frequency domain frame signal and a second frequency domain frame signal; determining a first power spectral density based on the first frequency domain frame signal, and determining a second power spectral density based on the second frequency domain frame signal; determining a cross-power spectral density based on the first frequency domain frame signal and the second frequency domain frame signal; calculating a coherence function based on the first power spectral density, the second power spectral density, and the cross-power spectral density; and extracting coherence coefficient characteristics based on the coherence function.

[0008] Optionally, a helmet wearing intelligent detection method further includes: when in an unworn state, continuously collecting a preset number of sets of first acceleration signals and second acceleration signals; calculating corresponding sets of coherence coefficient features and neck transfer function features based on the sets of first acceleration signals and second acceleration signals; obtaining multiple sets of wearing status results based on the sets of coherence coefficient features, neck transfer function features, and a pre-trained classifier; triggering a helmet audible and visual alarm and a vehicle speed limit reminder when the multiple sets of wearing status results meet the unworn condition; and determining that the multiple sets of wearing status results do not meet the unworn condition as an instantaneous signal abnormality, without triggering an alarm.

[0009] Optionally, determining the preset number of multiple sets of first acceleration signals and second acceleration signals includes: determining the signal fluctuation amplitude based on the first frequency band signal and the second frequency band signal; and determining the preset number based on the signal fluctuation amplitude, wherein the larger the signal fluctuation amplitude, the larger the preset number.

[0010] Optionally, determining the frequency band adjustment parameters includes: determining the road surface conditions at the current location in the previous target road segment using a map and a road condition database; classifying the road according to the road surface conditions to obtain the road surface classification at the current location; matching a preset frequency band adjustment reference value based on the road surface classification at the current location; obtaining the average driving speed of the previous target road segment; and calibrating the frequency band adjustment reference value based on the average driving speed of the previous target road segment to obtain the final frequency band adjustment parameters.

[0011] Optionally, the pre-trained classifier is a CNN-LSTM fusion classifier, whose training sample set includes the neck transfer function feature parameters and coherence coefficient features corresponding to helmet-wearing and helmet-free states under different road surface grades and different driving speeds.

[0012] According to a second aspect, the present invention provides a helmet-wearing intelligent detection device, comprising: a signal acquisition module for acquiring a first acceleration signal of a vehicle and a second acceleration signal of a helmet during vehicle operation; a frequency band selection module for selecting the same frequency band from the frequency domain signals of the first acceleration signal and the second acceleration signal based on a pre-set frequency band selection strategy, to obtain a first frequency band signal and a second frequency band signal, wherein the first frequency band signal and the second frequency band signal are frequency bands in which the energy proportion of the active motion component of the human head is lower than a preset threshold; a feature parameter determination module for determining feature parameters of the neck transfer function based on the first frequency band signal and the second frequency band signal; a coherence coefficient determination module for determining the coherence coefficient feature of the first acceleration signal and the second acceleration signal based on the components of the first acceleration signal and the second acceleration signal in the same frequency band; and a wearing judgment module for determining whether the helmet is in a wearing state based on the feature parameters of the neck transfer function, the coherence coefficient feature, and a pre-trained classifier.

[0013] According to a third aspect, an embodiment of the present invention provides an electronic device, the device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor performs the steps of the intelligent helmet wearing detection method described in the first aspect or any embodiment of the first aspect.

[0014] According to a fourth aspect, embodiments of the present invention provide a computer storage medium storing computer instructions thereon, which, when executed by a processor, implement the steps of a helmet-wearing intelligent detection method as described in the first aspect or any embodiment of the first aspect.

[0015] This invention provides an intelligent helmet wearing detection method. By collecting acceleration signals from the vehicle and the helmet, the method filters out frequency bands with low interference from active head movements. It then combines the neck transfer function characteristic parameters, coherence coefficient characteristics, and a pre-trained classifier to determine the wearing status. By utilizing the vibration transmission physical characteristics of the vehicle, neck, and helmet, the method distinguishes between a helmet actually being worn and one being placed on the vehicle. This effectively solves the problem of high false positive rates on bumpy roads and improves the accuracy and reliability of helmet wearing detection.

[0016] Other advantages, objectives, and features of the invention will be set forth in the following description and will be apparent to those skilled in the art in some respects, or may be learned by practice of the invention. The objectives and other advantages of the invention can be realized and obtained through the following description. Attached Figure Description

[0017] To make the objectives, technical solutions, and beneficial effects of this invention clearer, the following figures are provided for illustration: Figure 1 This is a flowchart illustrating a specific example of an intelligent helmet wearing detection method according to the present invention; Figure 2 This is a schematic diagram of a module structure of a helmet-wearing intelligent detection device according to the present invention; Figure 3 This is a schematic block diagram of a specific example of an electronic device in an embodiment 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, not all, of the embodiments of the present invention. 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.

[0019] 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; they can also refer to the internal connection of two components; and they can refer to a wireless connection or a wired connection. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0020] Furthermore, the technical features involved in the different embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0021] This invention provides a method for intelligent detection of helmet wearing, such as... Figure 1 As shown, it includes: S101, acquire the first acceleration signal of the vehicle and the second acceleration signal of the helmet during the vehicle's movement; S102, based on a pre-set frequency band selection strategy, select the same frequency band from the frequency domain signals of the first acceleration signal and the second acceleration signal to obtain the first frequency band signal and the second frequency band signal. The first frequency band signal and the second frequency band signal are frequency bands where the energy ratio of the active motion component of the human head is lower than a preset threshold. S103, Based on the first frequency band signal and the second frequency band signal, determine the characteristic parameters of the neck transfer function; S104, Based on the components of the first acceleration signal and the second acceleration signal in the same frequency band, determine the coherence coefficient characteristics of the two signals; S105 determines whether a helmet is being worn based on the characteristic parameters of the neck transfer function, the coherence coefficient, and a pre-trained classifier.

[0022] For example, some helmet-wearing recognition methods use the coherence coefficient to directly determine the wearing status. However, if the user loosely secures the helmet to the vehicle, the coherence coefficient may fall within the range of values ​​for the wearing status, making it indistinguishable from the actual wearing status. Therefore, the coherence coefficient cannot identify this type of false wearing. Thus, this embodiment introduces the characteristic parameters of the neck transfer function. Combining the two, they complement each other, helping to identify false wearing and further improving accuracy. The specific process is as follows: A first accelerometer sensor is installed at a predetermined location on the vehicle body, such as under the handlebars or seat. A second accelerometer sensor is installed at a predetermined location on the inside of the smart helmet, such as the inside of the forehead. When the vehicle is in motion, the first acceleration signal of the vehicle and the second acceleration signal of the helmet are collected simultaneously. Then, a Fast Fourier Transform is performed on the collected first and second acceleration signals to obtain their frequency domain signals. Based on a pre-set frequency band selection strategy, components are selected from a specific frequency band to obtain the first frequency band signal and the second frequency band signal. This frequency band can be determined through historical experiments, such as 2-20Hz. In this frequency band, the proportion of energy generated by active head movement is very low, that is, the proportion of active head movement energy in the selected frequency band is lower than a preset threshold, while passive vibration caused by road surface excitation during vehicle movement is dominant. The preset threshold value can be 10%. By selecting this frequency band, interference from active head movement can be filtered out, and the signal components that best reflect the transmission characteristics of the vehicle, neck, and helmet system can be extracted.

[0023] It is understandable that the neck transfer function characterizes the transmission of vehicle vibrations from the neck to the head. Specifically, when no helmet is worn, there is no effective coupling between the helmet and the head, and the characteristics of the neck transfer function change significantly. When a helmet is worn, the helmet and head form a unified whole, and the neck transfer function exhibits stable characteristics. Therefore, this embodiment calculates the characteristic parameters of the neck transfer function based on selected first and second frequency band signals. Specifically, the first frequency band signal of the vehicle can be used as input, and the second frequency band signal of the helmet can be used as output. The neck transfer function is estimated using the frequency response function.

[0024] Specifically, the average periodogram method is used to estimate the self-power spectral density of the vehicle-side signal and the cross-power spectral density between them. This is based on the self-power spectral density characterizing the energy distribution of the vehicle signal. and the cross-power spectral density between the two The frequency response function is obtained. Next, characteristic parameters are extracted from the frequency response function. Since the human neck is a multi-degree-of-freedom vibration system, there will be specific resonance peaks when wearing a device; when not wearing a device, there are no obvious resonance peaks or they coincide with the resonance peaks of the vehicle suspension system. Therefore, the resonance peak frequency can be used as one of the characteristic parameters. At the same time, the neck has a significant isolation effect on high-frequency vibrations and a large attenuation slope; rigid connections have a small attenuation slope. Therefore, the amplitude attenuation slope can also be used as one of the characteristic parameters.

[0025] The coherence coefficient characterizes the degree of linear correlation between the vehicle-side and helmet-side acceleration signals within the same effective frequency band. When the helmet is worn, the vibration transmission of the two signals shows a strong correlation, resulting in a high coherence coefficient value. When the helmet is not worn, the correlation between the two signals decreases significantly, resulting in a low coherence coefficient value. The specific determination process is as follows: Inverse Fourier transform is performed on the first and second frequency band signals to convert the frequency domain signals back to the time domain, obtaining the first frequency band time domain signal and the second frequency band time domain signal; the two time domain signals are then framed to obtain multiple continuous and overlapping first and second acceleration time frames, avoiding the omission of signal features; Fourier transform is performed on the first and second frequency band time domain signals within each time frame to obtain the corresponding first and second frequency domain frame signals; the first power spectral density is calculated based on the frequency domain frame signals. (That is, the self-power spectral density characterizing the energy distribution of vehicle signals), and the second power spectral density. That is, the self-power spectral density and cross-power spectral density characterizing the signal energy distribution on the helmet side. Calculate the coherence function using the following formula:

[0026] in, The coherence function has a value range of [0,1], with values ​​closer to 1 indicating stronger correlation. Features of the coherence function are extracted as coherence coefficient features, which can be the mean, maximum, minimum, and variance of the coherence function.

[0027] Finally, the neck transfer function feature parameters and coherence coefficient features are combined into a feature vector, which is then input into a pre-trained classifier. This classifier has learned the mapping relationship between features and wearing status by studying a large amount of sample data under different states. The classifier's output can accurately determine whether the helmet is currently in a true wearing state. It should be noted that the classifier can be a support vector machine (SVM), a neural network, etc., and this embodiment does not limit it. Those skilled in the art can determine it as needed.

[0028] This invention provides an intelligent helmet wearing detection method. By collecting acceleration signals from the vehicle and the helmet, the method filters out frequency bands with low interference from active head movements. It then combines the neck transfer function characteristic parameters, coherence coefficient characteristics, and a pre-trained classifier to determine the wearing status. By utilizing the vibration transmission physical characteristics of the vehicle, neck, and helmet, the method distinguishes between a helmet actually being worn and one being placed on the vehicle. This effectively solves the problem of high false positive rates on bumpy roads and improves the accuracy and reliability of helmet wearing detection.

[0029] As an optional implementation, based on a pre-set frequency band selection strategy, the same frequency band is selected from the frequency domain signals of the first acceleration signal and the second acceleration signal to obtain a first frequency band signal and a second frequency band signal. This includes: obtaining a target frequency band based on the vehicle's current position, wherein the target frequency band is predetermined based on the road surface characteristics at the current position when the vehicle was on the previous target road segment, and the previous target road segment is the road segment that the vehicle traversed during the target time period before the current position; and selecting the target frequency band from both the frequency domain signals of the first acceleration signal and the second acceleration signal to obtain the first frequency band signal and the second frequency band signal.

[0030] For example, this method obtains the vehicle's current location using an in-vehicle navigation system or a high-precision map while the vehicle is in motion. Based on this location, a pre-stored road condition database is queried. This database contains detailed road surface feature information for different road segments, such as road surface material, smoothness, and roughness. While traveling on the previous target road segment, that is, a short distance before the vehicle enters the current location, such as the first 10 meters, the system has already pre-determined a target frequency band suitable for the upcoming current location based on the road surface features of that segment.

[0031] As an optional approach, this embodiment pre-stores a road surface feature-frequency band mapping table in the cloud. This mapping table is pre-calibrated through extensive real-vehicle road tests and simulation analysis. For different types and grades of road surfaces, the vibration energy generated by the excitation vehicle exhibits different distribution characteristics in the frequency domain. To ensure that the proportion of active head motion energy in the selected frequency band is below a preset threshold, the analysis frequency band needs to be dynamically adjusted based on the dominant frequency range of the road surface excitation. Specifically, based on the geographical coordinates of the previous target road segment, the corresponding road surface features are retrieved from the road condition database. Then, based on the correspondence between the road surface features and the frequency band mapping table, the target frequency band under that road surface feature is determined.

[0032] For example, when the system detects that a vehicle is about to enter a rough gravel road from a smooth asphalt road, it automatically switches the target frequency band from a lower frequency band to a slightly higher frequency band based on the typical excitation spectrum of the gravel road. This is because rough road surfaces may contain more complex non-stationary components in the low-frequency band, while a specific high-frequency band better reflects the system characteristics. When the vehicle is actually driving on the gravel road, the method proposed in this embodiment directly filters and selects the frequency domain signals of the first and second acceleration signals in the pre-selected target frequency band to obtain the first frequency band signal and the second frequency band signal.

[0033] This invention provides an intelligent helmet wearing detection method. Based on the vehicle's current position, it matches a pre-determined target frequency band of the previous target road segment to achieve dynamic adaptation of the detection frequency band. This ensures that the selected frequency band always matches the vibration excitation characteristics of the current road surface, enabling more accurate extraction of effective signal components reflecting the wearing status. It further filters out interference from the road environment and improves the adaptability and accuracy of detection under different road conditions.

[0034] As an optional implementation, the coherence coefficient characteristics of the first and second acceleration signals are determined based on their components within the same frequency band, including: The first and second acceleration signals are subjected to frequency band filtering and inverse transformation to obtain time-domain signals in the first and second frequency bands. These signals are then framed to obtain multiple first and second acceleration time frames. The first and second frequency band time signals within each time frame are transformed to the frequency domain to obtain first and second frequency domain frame signals. Based on the first frequency domain frame signal, a first power spectral density is determined, and based on the second frequency domain frame signal, a second power spectral density is determined. Based on the first and second frequency domain frame signals, a cross-power spectral density is determined. Based on the first, second, and cross-power spectral densities, a coherence function is calculated. Based on the coherence function, coherence coefficient features are extracted.

[0035] For example, firstly, the original first and second acceleration signals are bandpass filtered (i.e., frequency band filtering) to retain components within the target frequency band. Then, they are converted back to the time domain using an inverse Fourier transform to obtain the first and second frequency band time domain signals. Since changes in road conditions and driving behavior during vehicle operation cause rapid changes in the statistical characteristics of the vibration signal over time, short-time analysis of the signal is necessary. If coherence function calculations are performed directly on the entire long-term signal without framing, the characteristics of transient events will be averaged out, failing to accurately reflect the true state of helmet wearing. Therefore, framing processing is performed on these two time domain signals. For example, a frame length of 256 sampling points and a frame shift of 128 sampling points are set to obtain multiple consecutive, partially overlapping first acceleration time frames and corresponding second acceleration time frames. This framing processing allows the analysis to capture the changes in the signal over time.

[0036] For each pair of time frames, they are again transformed to the frequency domain using FFT to obtain the first frequency domain frame signal and the second frequency domain frame signal. Then, the auto-power spectral density and cross-power spectral density of each frame are calculated. The auto-power spectral density can be obtained by dividing the square of the amplitude of the frequency domain frame signal by the frame length, and the cross-power spectral density is the product of the conjugates of the first and second frequency domain frame signals. The process of calculating the coherence function based on the first power spectral density, the second power spectral density, and the cross-power spectral density is described in the above embodiment and will not be repeated in this embodiment.

[0037] Finally, from the overall coherence function The quantized feature parameters are extracted as coherence coefficient features. These extracted features must comprehensively characterize the correlation degree of the signal. Specifically, through statistical analysis of the coherence function of each frame, statistical features that comprehensively characterize the time-varying characteristics of the signal correlation degree are extracted, including the mean, maximum and minimum values, and variance of the coherence function. These statistical features depict the distribution pattern and fluctuation characteristics of the coherence coefficient over time from different dimensions. Specifically, the mean of the coherence function reflects the overall correlation degree, the maximum and minimum values ​​reflect extreme cases, and the variance reflects the volatility of the correlation degree. After normalizing the extracted parameters, the final coherence coefficient feature vector is formed.

[0038] This invention provides an intelligent helmet wearing detection method. By calculating coherence coefficient features through frequency band filtering, inverse transformation, and frame processing, it can comprehensively and accurately characterize the correlation between vehicle and helmet acceleration signals, providing more accurate and effective feature basis for judging wearing status and reducing detection deviation caused by feature extraction errors.

[0039] As an optional implementation, a helmet wearing intelligent detection method further includes: when in an unworn state, continuously collecting a preset number of multiple sets of first acceleration signals and second acceleration signals; calculating corresponding multiple sets of coherence coefficient features and neck transfer function features based on the multiple sets of first acceleration signals and second acceleration signals; obtaining multiple sets of wearing status results based on the multiple sets of coherence coefficient features, neck transfer function features, and a pre-trained classifier; triggering a helmet audible and visual alarm and a vehicle speed limit reminder when the multiple sets of wearing status results meet the unworn condition; and determining that the multiple sets of wearing status results do not meet the unworn condition as an instantaneous signal abnormality, without triggering an alarm.

[0040] For example, if the initial wearing status result output by the classifier is an unworn state, the secondary verification process of this embodiment is triggered. Specifically, multiple sets of acceleration signals are continuously collected, keeping the sampling frequency and synchronization of the acceleration sensor unchanged, and a preset number of K sets of first acceleration signals and second acceleration signals are continuously collected. The collection duration of each set of signals is consistent with the collection duration of the initial detection, ensuring that the feature extraction conditions of multiple sets of signals are consistent.

[0041] For the collected K groups of first and second acceleration signals, the steps of the first embodiment are followed in sequence to select the frequency band, extract the neck transfer function feature parameters, and extract the coherence coefficient features to obtain the neck transfer function feature parameters and coherence coefficient feature fusion feature vectors corresponding to K groups; each group of fused feature vectors is input into a pre-trained classifier to obtain the K groups of wearing status results.

[0042] Pre-setting the non-wearing condition allows you to determine the percentage of non-wearing results in the K groups. A preset ratio, such as 80%, is set. When the condition of not wearing the helmet is met, it is determined that the helmet is truly not being worn, and the helmet's audible and visual alarm and vehicle speed limit reminder are immediately triggered. The helmet has a built-in buzzer and LED warning light. The buzzer emits a continuous alarm sound, and the LED light flashes red. When the condition of not wearing the helmet is not met, that is, when it is determined to be a momentary signal abnormality, such as a brief bump in the vehicle or a sudden change in characteristics caused by occasional shaking of the helmet, the helmet is actually being worn, and no alarm or speed limit measures are triggered. The system returns to the normal real-time detection process.

[0043] If the alarm is triggered and the vehicle occupants are wearing helmets, the detection method will collect signals in real time and re-extract features. Once the classifier outputs the wearing status result and the secondary verification passes, the helmet audible and visual alarm and vehicle speed limit will be immediately deactivated, and the vehicle will resume normal driving.

[0044] This invention provides an intelligent helmet wearing detection method. After the initial determination that the helmet is not being worn, multiple sets of signal secondary verification processes are added. By comprehensively judging the wearing status through the analysis results of multiple sets of features, it can effectively identify instantaneous signal anomalies caused by brief vehicle bumps or occasional helmet shaking, avoiding false alarms caused by single signal interference, and ensuring the accuracy of detection. At the same time, it triggers audible and visual alarms and vehicle speed limit reminders, which can promptly urge drivers and passengers to wear helmets properly and improve the safety of vehicle driving.

[0045] As an optional implementation, determining the preset number of multiple sets of first acceleration signals and second acceleration signals includes: Based on the first frequency band signal and the second frequency band signal, the signal fluctuation amplitude is determined; based on the signal fluctuation amplitude, a preset quantity is determined, wherein the larger the signal fluctuation amplitude, the larger the preset quantity.

[0046] For example, the standard deviations of the first frequency band signal and the second frequency band signal are first calculated to obtain the standard deviations of the first frequency band signal and the second frequency band signal. Then, the average of the two is calculated as the signal fluctuation amplitude S, which characterizes the overall signal fluctuation of both the vehicle and the helmet.

[0047] The optimal preset quantity K for different signal fluctuation ranges was determined through extensive real-vehicle testing. The core rule is that the larger the signal fluctuation amplitude, the larger the preset quantity K. A specific example of the correspondence is as follows: Low volatility range With K=3 groups, the signal fluctuation is small, and a small number of groups are sufficient to complete the effective verification. Mid-range fluctuation K=5 groups, signal fluctuation is moderate, increase the number of groups to avoid the influence of random fluctuations; High volatility range K=8 groups. When the signal fluctuates drastically, such as on gravel roads or bumpy roads, more groups are needed for accurate calibration. The correspondence between the above signal fluctuation amplitude range and the preset number K is stored in the vehicle controller's memory for real-time retrieval.

[0048] The calculated actual signal fluctuation amplitude S is matched with the preset fluctuation range, and the corresponding K value is retrieved as the number of acceleration signal groups continuously collected in this secondary verification.

[0049] In this embodiment, a larger signal fluctuation amplitude indicates a bumpier road condition and more interference components in the signal. In this case, the system automatically increases the preset number, for example, from the default 5 groups to 10 or 15 groups. Conversely, when the signal fluctuation amplitude is small, indicating a smooth road surface and a very stable signal, the system sets the preset number to a smaller number, such as 3 groups. This minimizes the confirmation time while ensuring reliable judgment, achieving a dynamic balance between detection accuracy and response speed.

[0050] As an optional implementation method, the determination of frequency band adjustment parameters includes: determining the road surface conditions at the current location in the previous target road segment using a map and a road condition database; classifying the road according to the road surface conditions to obtain the road surface classification at the current location; matching a preset frequency band adjustment reference value based on the road surface classification at the current location; obtaining the average driving speed of the previous target road segment; and calibrating the frequency band adjustment reference value based on the average driving speed of the previous target road segment to obtain the final frequency band adjustment parameters.

[0051] For example, firstly, in the previous target road segment, the system obtains the road surface conditions of that segment by fusing map data and a road condition database. The map can provide road classifications, while the road condition database can provide more detailed information, such as pavement type and road condition index.

[0052] Based on these road surface conditions, the system classifies roads. For example, road surfaces are classified into Grade 1 smooth asphalt road, Grade 2 rough asphalt road, Grade 3 cement road, Grade 4 gravel road, and Grade 5 severely potholed road, etc. Each road surface grade is pre-associated with a frequency band adjustment reference value, which is an empirical value or a frequency band center frequency or range calibrated through simulation. For example, for Grade 1 smooth asphalt road, the reference frequency band might be 5-15Hz; for Grade 4 gravel road, the reference frequency band might be 25-45Hz. The system matches the corresponding frequency band adjustment reference value based on the road surface grade at the current location.

[0053] Furthermore, considering that driving speed also affects the frequency components of road surface excitation—the faster the speed, the higher the vibration frequency generated by the same road surface—the system also acquires the vehicle's average driving speed on the previous target road segment. Then, based on this average speed, the matched frequency band adjustment reference value is calibrated. For example, a speed calibration coefficient table can be established, shifting the center frequency of the reference band upwards by 1 Hz for every 10 km / h increase in speed. Through speed calibration, the final, accurate frequency band adjustment parameters are obtained. These parameters will be used for frequency band selection on the next road segment.

[0054] This invention provides an intelligent helmet wearing detection method that combines road surface grading with frequency band adjustment benchmarks and calibrates benchmarks using the average driving speed of the previous target road segment to obtain precise frequency band adjustment parameters. This allows the detection frequency band to simultaneously adapt to the two key influencing factors of road surface conditions and driving speed, accurately locking the effective frequency bands that reflect the wearing status under different driving conditions, and further improving the accuracy and adaptability of detection under complex driving conditions.

[0055] As an optional implementation, the pre-trained classifier is a CNN-LSTM fusion classifier, whose training sample set includes the neck transfer function feature parameters and coherence coefficient features corresponding to helmet-wearing and helmet-free states under different road surface grades and different driving speeds.

[0056] For example, CNN excels at extracting local spatial features of signals and can effectively capture the local correlation patterns of neck transfer function features and coherence coefficient features; LSTM excels at processing time-series data and can effectively capture the temporal dynamic features of acceleration signals, such as the vibration of a vehicle and a helmet, which is a continuous time-series signal. In order to combine the advantages of both, this embodiment fuses CNN and LSTM to form a CNN-LSTM fusion classifier.

[0057] The CNN-LSTM network structure, from bottom to top, consists of an input layer, a CNN feature extraction layer, an LSTM temporal feature extraction layer, a fully connected layer, and an output layer. The input layer takes a fused feature vector as input, with a dimension of [missing information]. Where D is the feature dimension, which is the total number of neck transfer function feature parameters and coherence coefficient features, and T is the time step; the CNN feature extraction layer consists of two one-dimensional convolutional layers + pooling layers, with a convolutional kernel size of 3, a stride of 1, and the activation function ReLU. The pooling layer uses max pooling with a pooling kernel size of 2, used to extract local spatial features of the fused features and reduce the feature dimension; the LSTM temporal feature extraction layer consists of two LSTM layers, each with 64 neurons and a dropout rate of 0.2, used to extract the temporal dynamic features of the CNN output features and capture the temporal correlation of the signal; the fully connected layer consists of one fully connected layer with 32 neurons and the activation function ReLU, which fuses the features extracted by CNN and LSTM; the output layer consists of one fully connected layer with 2 neurons and the activation function Softmax, and the output is the probability value of wearing / not wearing.

[0058] The classifier training process includes: First, using the training set as input, the Adam optimizer is used for model training, with the cross-entropy loss function as the loss function. The training batch size is set to 32, the number of training epochs is set to 50, and the initial learning rate is 0.001, which decays to 0.5 every 10 epochs. The sample set construction process involves: selecting combinations of different road surface grades and different driving speeds; collecting samples in both helmet-wearing and helmet-free states under each combination; collecting the vehicle's first acceleration signal and the helmet's second acceleration signal under each condition; extracting the neck transfer function feature parameters and coherence coefficient features; fusing these features to form a sample feature vector; manually labeling each sample feature vector, marking the wearing state as "1" and the helmet-free state as "0"; and dividing the collected and labeled sample set into a training set, validation set, and test set in a 7:2:1 ratio.

[0059] Next, the hyperparameters of the model, such as the number of convolutional kernels, the number of LSTM neurons, and the dropout rate, are fine-tuned based on the validation set. The hyperparameter combination that achieves the highest classification accuracy on the validation set is selected. Finally, the test set is input into the trained classifier to verify the model's classification accuracy, requiring a certain classification accuracy on the test set. If the target is not reached, the sample size is increased and the model is retrained. The trained and validated CNN-LSTM fusion classifier model is saved to the vehicle controller's memory for use during real-time detection.

[0060] This invention provides an intelligent helmet wearing detection method that employs a CNN-LSTM fusion classifier. It combines the advantages of CNN in extracting local spatial features and LSTM in capturing temporal dynamic features. Furthermore, the training samples cover wearing states under different road surface grades and driving speeds, enabling the classifier to accurately identify feature patterns under different working conditions and improve the accuracy of wearing state judgment. At the same time, it integrates the neck transfer function and coherence coefficient as dual features, making the classifier's judgment basis more comprehensive and further reducing the probability of misjudgment and missed judgment.

[0061] This embodiment provides a smart helmet-wearing detection device, such as... Figure 2 As shown, it includes: The signal acquisition module 201 is used to acquire the first acceleration signal of the vehicle and the second acceleration signal of the helmet during the vehicle's movement. The frequency band selection module 202 is used to select the same frequency band from the frequency domain signal of the first acceleration signal and the frequency domain signal of the second acceleration signal based on a preset frequency band selection strategy, so as to obtain a first frequency band signal and a second frequency band signal. The first frequency band signal and the second frequency band signal are frequency bands in which the energy ratio of the active motion component of the human head is lower than a preset threshold. The characteristic parameter determination module 203 is used to determine the characteristic parameters of the neck transfer function based on the first frequency band signal and the second frequency band signal; The coherence coefficient determination module 204 is used to determine the coherence coefficient characteristics of the first acceleration signal and the second acceleration signal based on the components of the first acceleration signal and the second acceleration signal in the same frequency band. Wearing determination module 205 is used to determine whether the helmet is being worn based on the feature parameters of the neck transfer function, the coherence coefficient feature, and a pre-trained classifier.

[0062] This application also provides an electronic device, such as... Figure 3 As shown, processor 501 and memory 502 are connected via a bus or other means.

[0063] Processor 501 can be a central processing unit (CPU). Processor 501 can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, or combinations of the above types of chips.

[0064] The memory 502, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs, non-transitory computer-executable programs, and modules, such as the program instructions / modules corresponding to a helmet-wearing intelligent detection method in this embodiment of the invention. The processor executes various functional applications and data processing by running the non-transitory software programs, instructions, and modules stored in the memory.

[0065] Memory 502 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created by the processor, etc. Furthermore, the memory may include high-speed random access memory and non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory 502 may optionally include memory remotely located relative to the processor, which can be connected to the processor via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0066] The one or more modules are stored in the memory 502, and when executed by the processor 501, they perform actions such as... Figure 1 An intelligent helmet-wearing detection method is shown in the embodiment.

[0067] For specific details regarding the aforementioned electronic devices, please refer to the relevant documentation. Figure 1 The relevant descriptions and effects in the illustrated embodiments are for understanding purposes only and will not be repeated here.

[0068] This embodiment also provides a computer storage medium storing computer-executable instructions that can execute a helmet-wearing intelligent detection method in any of the above method embodiments. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk drive (HDD), or solid-state drive (SSD), etc.; the storage medium may also include combinations of the above types of memory.

[0069] Finally, it should be noted that the above preferred embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail through the above preferred embodiments, those skilled in the art should understand that various changes can be made to it in form and detail without departing from the scope defined by the claims of the present invention.

Claims

1. A method for intelligent detection of helmet wearing, characterized in that, include: Acquire the vehicle's first acceleration signal and the helmet's second acceleration signal during vehicle operation; Based on a pre-set frequency band selection strategy, the same frequency band is selected from the frequency domain signals of the first acceleration signal and the second acceleration signal to obtain the first frequency band signal and the second frequency band signal. The first frequency band signal and the second frequency band signal are the frequency bands where the energy proportion of the active motion component of the human head is lower than a preset threshold. Based on the first frequency band signal and the second frequency band signal, the characteristic parameters of the neck transfer function are determined; Based on the components of the first acceleration signal and the second acceleration signal in the same frequency band, the characteristics of their coherence coefficients are determined. Based on the characteristic parameters of the neck transfer function, the coherence coefficient, and a pre-trained classifier, it is determined whether the helmet is being worn.

2. The intelligent helmet wearing detection method according to claim 1, characterized in that, Based on a pre-defined frequency band selection strategy, the same frequency band is selected from the frequency domain signals of the first acceleration signal and the second acceleration signal to obtain a first frequency band signal and a second frequency band signal, including: Based on the vehicle's current location, the target frequency band is obtained. The target frequency band is determined in advance based on the road surface characteristics of the current location when the vehicle was in the previous target road segment. The previous target road segment is the road segment that the vehicle passed through in the target time period before the current location. Target frequency bands are selected from both the frequency domain signals of the first acceleration signal and the frequency domain signals of the second acceleration signal to obtain the first frequency band signal and the second frequency band signal.

3. The intelligent helmet wearing detection method according to claim 1, characterized in that, Based on the components of the first and second acceleration signals within the same frequency band, the characteristics of their coherence coefficients are determined, including: The first acceleration signal and the second acceleration signal are subjected to frequency band filtering and inverse transformation to obtain the time domain signal of the first frequency band and the time domain signal of the second frequency band. The time-domain signals of the first frequency band and the second frequency band are processed by frame segmentation to obtain multiple first acceleration time frames and multiple second acceleration time frames. Transform the first frequency band time domain signal and the second frequency band time domain signal in each time frame to the frequency domain to obtain the first frequency domain frame signal and the second frequency domain frame signal. The first power spectral density is determined based on the first frequency domain frame signal, and the second power spectral density is determined based on the second frequency domain frame signal. The cross-power spectral density is determined based on the first frequency domain frame signal and the second frequency domain frame signal. Calculate the coherence function based on the first power spectral density, the second power spectral density, and the cross-power spectral density; Based on the coherence function, the coherence coefficient features are extracted.

4. The intelligent helmet wearing detection method according to claim 1, characterized in that, Also includes: When not wearing the device, a preset number of sets of first and second acceleration signals are continuously collected. Based on multiple sets of first and second acceleration signals, the corresponding multiple sets of coherence coefficient characteristics and neck transfer function characteristics are calculated respectively. Based on multiple sets of coherence coefficient features, neck transfer function features, and a pre-trained classifier, multiple sets of wearing status results are obtained; If multiple wearing status results meet the condition of not wearing, the helmet's audio and visual alarm and vehicle speed limit reminder will be triggered. If multiple sets of wearing status results do not meet the condition of not wearing, it is determined to be an instantaneous signal abnormality and no alarm is triggered.

5. The intelligent helmet wearing detection method according to claim 4, characterized in that, The determination of the preset number of multiple sets of first acceleration signals and second acceleration signals includes: The signal fluctuation amplitude is determined based on the first frequency band signal and the second frequency band signal. Based on the signal fluctuation amplitude, a preset quantity is determined, where the larger the signal fluctuation amplitude, the larger the preset quantity.

6. The intelligent helmet wearing detection method according to claim 2, characterized in that, Determining the frequency band adjustment parameters includes: In the previous target road segment, the road surface conditions at the current location are determined using the map and road condition database; Based on the road surface conditions, the road is classified to obtain the road surface classification at the current location; Based on the road surface classification at the current location, the corresponding preset frequency band adjustment reference value is used; Get the average speed of the previous target road segment; Based on the average driving speed of the previous target road segment, the frequency band adjustment reference value is calibrated to obtain the final frequency band adjustment parameters.

7. A helmet-wearing intelligent detection method according to any one of claims 1-6, characterized in that, The pre-trained classifier is a CNN-LSTM fusion classifier. Its training sample set includes the neck transfer function feature parameters and coherence coefficient features corresponding to helmet-wearing and helmet-free states under different road surface grades and different driving speeds.

8. A helmet-wearing intelligent detection device, characterized in that, include: The signal acquisition module is used to acquire the vehicle's first acceleration signal and the helmet's second acceleration signal during vehicle operation. The frequency band selection module is used to select the same frequency band from the frequency domain signals of the first acceleration signal and the second acceleration signal based on a preset frequency band selection strategy, so as to obtain the first frequency band signal and the second frequency band signal. The first frequency band signal and the second frequency band signal are frequency bands where the energy ratio of the active motion component of the human head is lower than a preset threshold. The characteristic parameter determination module is used to determine the characteristic parameters of the neck transfer function based on the first frequency band signal and the second frequency band signal; The coherence coefficient determination module is used to determine the coherence coefficient characteristics of the first acceleration signal and the second acceleration signal based on the components of the first acceleration signal and the second acceleration signal in the same frequency band. The helmet wearing determination module is used to determine whether the helmet is being worn based on the feature parameters of the neck transfer function, the coherence coefficient features, and a pre-trained classifier.

9. An electronic device, the device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor performs the steps of the intelligent helmet-wearing detection method according to any one of claims 1-7.

10. A computer storage medium storing computer instructions thereon, characterized in that, When executed by the processor, this instruction implements the steps of the intelligent helmet wearing detection method according to any one of claims 1-7.

Images