Hybrid sensor positioning system for pedestrian step length estimation

Abstract

The present application provides a hybrid sensor positioning system for pedestrian stride length estimation, which relates to the technical field of stride length recognition. The method includes: a connection establishment unit, which is used to establish a communication connection with an intelligent sensor module; a human feature acquisition unit, which is used to obtain target human features; a signal noise reduction unit, which is used to generate stable sensor signal data; a modal matching unit, which is used to obtain modal matching results; and a stride length recognition unit, which is used to generate stride length estimation results. The present application can solve the technical problem in the prior art that the stride length recognition accuracy is low due to the difficulty in performing stride length recognition for different features of the left and right feet. By performing stride length recognition modal matching after performing noise reduction processing on the sensor data, stride length recognition under different modalities can be realized, thereby achieving the technical effect of improving the efficiency and accuracy of stride length recognition.

Description

Technical Field

[0001] The present application relates to the technical field of stride length recognition, and in particular to a hybrid sensor positioning system for pedestrian stride length estimation. Background Art

[0002] Nowadays, more and more intelligent services require location services as a prerequisite. With the rapid development of location-based service applications and the gradual maturity of micro-electromechanical system technology, sensor-based pedestrian sensing navigation has become a research hotspot because of its wide range of applications and the ability to achieve autonomous navigation without relying on external information.

[0003] Currently, existing pedestrian stride length recognition mostly uses pedestrian dead reckoning algorithms for stride length recognition. However, different pedestrians have different physical characteristics, and the stride lengths of pedestrians' left and right feet may be inconsistent. Existing technology is difficult to perform stride length recognition based on the different characteristics of the left and right feet, and the recognition results are subject to errors.

[0004] In summary, the prior art has a technical problem in that it is difficult to identify stride lengths based on the different features of the left and right feet, resulting in low stride length recognition accuracy. Summary of the Invention

[0005] The purpose of this application is to provide a hybrid sensor positioning system for pedestrian stride length estimation, so as to solve the technical problem in the prior art that it is difficult to identify stride length based on the different features of the left and right feet, resulting in low stride length recognition accuracy.

[0006] In view of the above problems, the present application provides a hybrid sensor positioning system for pedestrian stride length estimation.

[0007] In the first aspect, the present application provides a hybrid sensor positioning system for pedestrian stride length estimation, the system comprising: a connection establishment unit, the connection establishment unit being used to establish a communication connection with an intelligent sensing module, wherein the intelligent sensing module is embedded in an intelligent terminal; a human body feature acquisition unit, the human body feature acquisition unit being used to collect human body features of a target pedestrian to obtain target human body features; a signal noise reduction unit, the signal noise reduction unit being used to receive the sensing signal data collected by the intelligent sensing module and input it into a noise reduction processing network to perform signal noise reduction and generate stable sensing signal data; a modal matching unit, the modal matching unit being used to perform two-foot balance state recognition based on the stable sensing signal data, and perform stride length recognition modal matching according to the balance state recognition result to obtain a modal matching result, wherein the stride length recognition modality includes a single-step mode or a double-step mode; a stride length recognition unit, the stride length recognition unit being used to analyze the stable sensing signal data and the target human body features according to the modal matching result through a stride length recognition model to generate a stride length estimation result.

[0008] In a second aspect, the present application provides a hybrid sensor positioning method for pedestrian stride length estimation, which is implemented by a hybrid sensor positioning system for pedestrian stride length estimation, wherein the method includes: establishing a communication connection with an intelligent sensing module, wherein the intelligent sensing module is embedded in an intelligent terminal; collecting human body features of a target pedestrian to obtain target human body features; receiving the sensor signal data collected by the intelligent sensing module and inputting it into a noise reduction processing network to perform signal noise reduction to generate stable sensor signal data; performing double-foot balance state recognition based on the stable sensor signal data, and performing stride length recognition modal matching according to the balance state recognition result to obtain a modal matching result, wherein the stride length recognition modal includes a single-step mode or a double-step mode; according to the modal matching result, the stable sensor signal data and the target human body features are analyzed through a stride length recognition model to generate a stride length estimation result.

[0009] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0010] Establishing a communication connection with an intelligent sensing module, wherein the intelligent sensing module is embedded in an intelligent terminal; collecting human body features of a target pedestrian to obtain target human body features; receiving the sensor signal data collected by the intelligent sensing module and inputting it into a noise reduction processing network to perform signal noise reduction to generate stable sensor signal data; performing two-leg balance state recognition based on the stable sensor signal data, performing stride length recognition modal matching according to the balance state recognition result, and obtaining a modal matching result, wherein the stride length recognition modality includes a single-step mode or a double-step mode; according to the modal matching result, analyzing the stable sensor signal data and the target human body features through a stride length recognition model to generate a stride length estimation result. Thus, by performing noise reduction processing on the sensor data and then performing stride length recognition modal matching, stride length recognition is performed according to the modal matching result, combined with the sensor data and human body features, to achieve stride length recognition under different modalities, thereby achieving the technical effect of improving the efficiency and accuracy of stride length recognition.

[0011] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, which can be implemented in accordance with the contents of the description, and to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are specifically listed below. It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present application, nor is it intended to limit the scope of the present application. Other features of the present application will become easy to understand through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] In order to more clearly illustrate the technical solutions in this application or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are merely exemplary, and a person of ordinary skill in the art can obtain other drawings based on the provided drawings without creative work.

[0013] Figure 1 A flow chart of a hybrid sensor positioning method for pedestrian step length estimation in this application;

[0014] Figure 2 This is a schematic diagram of the structure of the hybrid sensor positioning system used for pedestrian stride length estimation in this application.

[0015] Explanation of the accompanying symbols: connection establishment unit 11, human feature acquisition unit 12, signal noise reduction unit 13, modality matching unit 14, step length recognition unit 15. DETAILED DESCRIPTION

[0016] This application provides a hybrid sensor positioning system for pedestrian stride length estimation, solving the technical problem of low stride length recognition accuracy in the prior art due to the difficulty in identifying stride length based on the different features of the left and right feet. By performing stride length recognition modal matching after noise reduction processing on the sensor data, stride length recognition is performed based on the modal matching results, combining the sensor data with human body characteristics, and achieving stride length recognition in different modalities, thereby achieving the technical effect of improving the efficiency and accuracy of stride length recognition.

[0017] Below, the technical solutions in this application will be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of this application, rather than all the embodiments of this application. It should be understood that this application is not limited to the example embodiments described herein. Based on the embodiments of this application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application. It should also be noted that, for the convenience of description, only the parts related to this application, rather than all of them, are shown in the accompanying drawings.

[0018] Example 1

[0019] Please see the attached Figure 1 The present application provides a hybrid sensor positioning method for pedestrian stride length estimation, wherein the method is applied to a hybrid sensor positioning system for pedestrian stride length estimation, and the method specifically comprises the following steps:

[0020] Step 1: Establishing a communication connection with an intelligent sensor module, wherein the intelligent sensor module is embedded in an intelligent terminal;

[0021] Specifically, the intelligent sensing module refers to a sensor integrated module used in the prior art to detect the real-time motion status of pedestrians. The intelligent sensing module includes at least an accelerometer, a gyroscope, and a magnetometer. The accelerometer, the gyroscope, and the magnetometer are packaged and embedded in the intelligent terminal. In simple terms, the accelerometer, the gyroscope, and the magnetometer are embedded in the same mobile device, such as a mobile phone. The mobile phone may include an accelerometer, a gyroscope, and a magnetometer. The intelligent terminal is generally worn by the target pedestrian, and can collect sensor signal data such as acceleration and angular velocity of the pedestrian when walking in real time. Therefore, a connection with the intelligent sensing module in the intelligent terminal is usually established through wireless connection methods such as WIFI and Bluetooth. This is a commonly used technical means for those skilled in the art and will not be expanded here.

[0022] Step 2: Collect human features of the target pedestrian to obtain the target human features;

[0023] Specifically, the target pedestrian refers to the pedestrian whose stride length is to be estimated. The target human body characteristics include the height, weight and gender of the target pedestrian, which can be input by the user through the user terminal.

[0024] Step 3: receiving the sensor signal data collected by the intelligent sensor module and inputting it into the noise reduction processing network to perform signal noise reduction and generate stable sensor signal data;

[0025] Specifically, the sensor signal data collected by the intelligent sensing module refers to data collected by the accelerometer, gyroscope, and magnetometer within the smart terminal. Because pedestrians are subject to noise interference during walking, the acceleration, angular velocity, and magnetic intensity data directly collected by the intelligent sensing module contain a high level of noise, necessitating noise reduction processing. The noise reduction processing network is a machine learning model built based on an autoencoder, which is used to perform noise reduction on the sensor signal data, obtaining the noise-reduced sensor signal data as stable sensor signal data. This improves the stability of the underlying data and, in turn, the accuracy of step length estimation.

[0026] Step 4: performing bipedal balance state recognition based on the stable sensor signal data, and performing stride length recognition mode matching according to the balance state recognition result to obtain a mode matching result, wherein the stride length recognition mode includes a single-step mode or a double-step mode;

[0027] Specifically, in the prior art, when performing stride length prediction, the distance between adjacent landing points of the same part of the two feet is used as the prediction result, that is, the stride lengths of the left foot and the right foot are identified separately. However, some pedestrians may have leg injuries, resulting in inconsistent stride lengths of the two feet. Then, based on the prediction results, the pedestrian's walking time, distance, etc. are analyzed. However, the stride lengths of different pedestrians may be consistent or inconsistent. For example, due to different pedestrians' different balance states, for example, some pedestrians may have leg injuries, resulting in inconsistent stride lengths of the two feet. If the stride lengths of the two feet are consistent, matching the single-step state, only single-step step length recognition is required, which can improve efficiency. If they are inconsistent, matching the double-step state, and performing stride length recognition on both feet, thus executing different stride length recognition modes for different states, it is convenient to improve the efficiency of stride length recognition while improving recognition accuracy.

[0028] Step 5: According to the modal matching result, the stable sensing signal data and the target human body characteristics are analyzed through a step length recognition model to generate a step length estimation result.

[0029] Specifically, the step length recognition model is constructed based on the existing machine learning model, and the step length is automatically recognized based on the stable sensor signal data and the target human body characteristics to obtain the step length estimation result, which includes the step length and the pace. Specifically, multiple groups of sensor signal samples and human body feature samples can be obtained through existing technology, and the corresponding step length and pace are obtained as step length estimation result samples. Then, based on the existing machine learning model, such as a neural network model, a step length recognition model is constructed, and the group of sensor signal samples and human body feature samples, as well as the corresponding step length estimation result samples are used as training data to train the step length recognition model. The step length recognition model is trained until convergence, and a step length recognition model that meets the requirements can be obtained. Then, the stable sensor signal data and the target human body characteristics are input into the step length recognition model, and the output result is the step length estimation result. The training of the machine learning model is a common technical means used by those skilled in the art and will not be expanded here.

[0030] Furthermore, step three of this application also includes:

[0031] Noise is added to the sensor signal data to obtain a damaged sensor signal; the noise reduction processing network is constructed based on an autoencoder, wherein the noise reduction processing network includes an encoder and a decoder; deep feature extraction is performed on the damaged sensor signal using the noise reduction processing network, and the original signal is reconstructed to obtain the stable sensor signal data.

[0032] Furthermore, the present application further comprises the following steps:

[0033] Acquire first noise-added data; acquire a historical sensor signal acquisition sample set, and perform noise fusion on the first noise-added data to obtain a first damaged sensor signal sample set; construct a signal reconstruction training function, and train the encoder and the decoder in combination with the first damaged sensor signal sample set and the historical sensor signal acquisition samples to generate the noise reduction processing network.

[0034] Furthermore, the present application further comprises the following steps:

[0035] Construct the signal reconstruction training function:

[0036]

[0037] Among them, L(x,y) is the signal reconstruction loss value, x i is the i-th damage sensing signal sample in the first damage sensing signal sample set, y i is a reconstructed sensing signal output after the encoder and the decoder reconstruct the original signal of the i-th damage sensing signal sample, and N is the total number of samples in the first damage sensing signal sample set.

[0038] Specifically, the method of receiving the sensor signal data collected by the intelligent sensor module and inputting it into the noise reduction processing network to perform signal noise reduction and generate stable sensor signal data is as follows:

[0039] Because pedestrians are subject to noise interference while walking, the sensor signal data contains a lot of noise. Common noise reduction methods, such as mean filtering and median filtering, do not provide robust noise reduction results. Therefore, we do not apply for noise reduction processing through machine learning models to improve stability. Specifically, we first add noise to the sensor signal data to simulate signal damage. The added noise can be Gaussian noise, random noise, impulse noise, etc., which can be selected by professionals in this field. The sensor signal data after adding noise is used as the damage sensor signal.

[0040] Furthermore, the noise reduction processing network is constructed based on an autoencoder, wherein the noise reduction processing network includes an encoder and a decoder. The noise reduction processing network is composed of multiple deep learning networks and requires training with a large amount of training data. The specific method is as follows: obtaining first noise-added data, wherein the first noise-added data is Gaussian noise, random noise, impulse noise, or other noise, without limitation, and the first noise-added data can include multiple noise data of different intensities, which are subsequently superimposed with historical sensor signal acquisition samples. Obtaining a historical sensor signal acquisition sample set, the historical sensor signal acquisition sample set refers to sensor signal data collected and obtained over a period of time. The first noise-added data is fused with each sample data in the first noise-added data to obtain a first damage sensor signal sample set.

[0041] Further, construct the signal reconstruction training function:

[0042] Among them, L(x,y) is the signal reconstruction loss value, x i is the i-th damage sensing signal sample in the first damage sensing signal sample set, y i is a reconstructed sensing signal output after the encoder and the decoder reconstruct the original signal of the i-th damage sensing signal sample, and N is the total number of samples in the first damage sensing signal sample set.

[0043] That is, the encoder and decoder need to be trained based on the first set of impaired sensor signal samples and the historical sensor signal samples. Reconstructing the sensor signal involves re-extracting deep features from the damaged signal to reconstruct the signal. That is, the i-th impaired sensor signal sample and the corresponding reconstructed sensor signal are not exactly the same, but they are similar. Therefore, during training, the difference between the two needs to be reduced. Specifically, the encoding weights and encoding biases of the encoder and decoder are adjusted to minimize the signal reconstruction loss. Consequently, when the signal reconstruction loss is less than or equal to a predetermined loss, the denoising network training is complete. This provides a model foundation for denoising.

[0044] Finally, the damage sensing signal is subjected to deep feature extraction using the noise reduction processing network, and the original signal is reconstructed to output the stable sensing signal data, thereby achieving stable noise reduction processing, improving the accuracy of basic data, and further improving the step length recognition accuracy.

[0045] Furthermore, step 4 of this application also includes:

[0046] Peaks and troughs are identified on the stable sensing signal data to obtain peak and trough time series data; single-step cutting is performed on the peak and trough time series data based on single-step cycle characteristics to obtain a continuous single-step feature sequence; single-step duration fluctuation analysis under continuous single-step cycles is performed on the continuous single-step feature sequence to obtain a double-step balance state index; if the double-step balance state index is greater than or equal to a predetermined balance index, the modal matching result is a double-step mode; if the double-step balance state index is less than the predetermined balance index, the modal matching result is a single-step mode.

[0047] Furthermore, the present application further comprises the following steps:

[0048] Acquire walking test sensor data samples and walking state timing samples; locate the step node sequence based on the walking state timing samples; perform data segmentation on the walking test sensor data samples based on the step node sequence to obtain multiple single-step cycle test data; perform peak and trough feature extraction on the multiple single-step cycle test data, and perform central value evaluation to obtain the single-step cycle feature.

[0049] Specifically, based on the stable sensor signal data, the balance state of both feet is identified, and the step length identification modal matching is performed according to the balance state identification result. The method for obtaining the modal matching result is as follows:

[0050] First, the stable sensor signal data is subjected to peak and trough identification, that is, the peak points and trough points in the stable sensor signal data are identified and arranged in chronological order, with the arrangement results being used as peak and trough time series data. The peak and trough time series data is further subjected to single-step segmentation based on the single-step periodicity feature, wherein the single-step periodicity feature refers to the peak and trough feature within a step length. For example, a large peak, a small peak, and a trough will appear within a step length, wherein the peak value of the large peak is greater than the peak value of the small peak. The specific value is difficult to accurately set. That is, during a step, the acceleration will reach a maximum value at a certain node and then decrease until it drops to a minimum value, i.e., a trough, and then start to rise again, but will not rise to the maximum value of the peak. At this point, it indicates that a step has been taken. Thus, by segmenting the peak and trough time series data according to the single-step periodicity feature, a continuous single-step feature sequence can be obtained. The continuous single-step feature sequence contains multiple continuous single-step sensor data.

[0051] Among them, the method for obtaining single-step cycle features is: obtaining walking test sensor data samples and walking state time series samples, the walking test sensor data samples are the monitoring data obtained by the tester carrying the intelligent sensor module to perform a walking test, and the walking state time series samples are the accurate time node sequence corresponding to each step of the tester, for example, the start of the first step corresponds to a time, the end of the first step corresponds to a time, the start of the second step corresponds to a time, and so on, to obtain the walking state time series samples. Then, based on the walking state time series samples, the step node sequence is located, that is, the time range of each step is located according to the start time and end time of each step to form a step node sequence. Further, based on the step node sequence, the walking test sensor data samples are segmented according to the time range corresponding to each step to obtain multiple single-step cycle test data corresponding to each step. Finally, peak and trough feature extraction is performed on the multiple single-step cycle test data, that is, the number of peaks and the number of troughs in the multiple single-step cycle test data are extracted respectively, as well as the size relationship between peaks and peaks, and troughs and troughs as multiple feature extraction results. Cluster analysis is then performed on the multiple feature extraction results to obtain multiple clusters. The number of clusters in the multiple clusters is then counted, and the feature extraction results within the cluster with the largest number of clusters are used as the single-step cycle features. This enables single-step cycle feature analysis, laying the foundation for subsequent analysis of the two-legged balance state.

[0052] Furthermore, a single-step duration fluctuation analysis is performed on the continuous single-step feature sequence under a continuous single-step cycle, that is, analyzing the duration difference of two adjacent single-step sensor data in the continuous single-step feature sequence. Specifically, the duration of any two adjacent single-step sensor data in the continuous single-step feature sequence can be extracted and then subtracted to obtain multiple subtraction results, thereby obtaining multiple subtraction results, and then performing fluctuation analysis on the multiple subtraction results. For example, the variance, standard deviation, and other indicators that can reflect the data fluctuation situation can be calculated as the balance state indicator of the two feet. Then, a predetermined balance indicator is set. The predetermined balance indicator refers to the time difference indicator when judging the consistency of the step length of the two feet, and is set by professionals in this field based on actual conditions.

[0053] If the double-foot balance state index is greater than or equal to the predetermined balance index, the modal matching result is a double-step mode, that is, at this time, the stride lengths of the left and right feet are inconsistent, so it is necessary to perform stride length identification and inference on both the left and right feet; if the double-foot balance state index is less than the predetermined balance index, the modal matching result is a single-step mode, indicating that the stride lengths of the left and right feet are basically the same, and it is only necessary to perform stride length identification and inference on a single foot, which can improve the efficiency and accuracy of stride length identification.

[0054] Furthermore, step five of this application also includes:

[0055] If the modal matching result is a two-step modality, periodic superposition is performed on the single-step periodic feature to obtain a two-step periodic feature; based on the two-step periodic feature, periodic data extraction is performed on the stable sensing signal data, and the two-step sensing periodic data and the target human body feature are input into the step length recognition model for analysis to generate the step length inference result; wherein, the step length inference result carries identification information including the modal matching result.

[0056] Specifically, if the modal matching result is a double-step modality, it is necessary to identify the left and right feet as a stride analysis unit, and it is necessary to extract the sensor data under one stride analysis unit. Therefore, the single-step periodic feature is subjected to periodic superposition, which is to double the features of the single-step periodic feature. For example, if the single-step periodic feature is two peaks and one trough, then the double-step periodic feature is four peaks and two troughs. Based on the double-step periodic feature, the stable sensor signal data is subjected to periodic data extraction, which is to segment the data corresponding to the number of peaks and troughs of the double-step periodic feature as double-step sensing periodic data. The double-step sensing periodic data and the target human body feature are input into the aforementioned stride recognition model for analysis, and the stride length estimation result can be output. The stride length estimation result at this time is the stride length and walking speed under the double-step condition. At the same time, the stride length estimation result carries the identification information of the modal matching result, which facilitates those skilled in the art to analyze the walking time or walking distance of the target pedestrian based on the stride length estimation result.

[0057] Similarly, if the modal matching result is a single-step mode, periodic data extraction is performed on the stable sensor signal data based on the single-step periodic characteristics. The single-step sensing periodic data and the target body characteristics are input into the stride length recognition model for analysis to generate the stride length estimation result. This improves the accuracy of stride length estimation by analyzing the balance state of the left and right feet and then estimating stride length in different modalities.

[0058] In summary, the hybrid sensor positioning method for pedestrian stride length estimation provided by this application has the following technical effects:

[0059] Establishing a communication connection with an intelligent sensing module, wherein the intelligent sensing module is embedded in an intelligent terminal; collecting human body features of a target pedestrian to obtain target human body features; receiving the sensor signal data collected by the intelligent sensing module and inputting it into a noise reduction processing network to perform signal noise reduction to generate stable sensor signal data; performing two-leg balance state recognition based on the stable sensor signal data, performing stride length recognition modal matching according to the balance state recognition result, and obtaining a modal matching result, wherein the stride length recognition modality includes a single-step mode or a double-step mode; according to the modal matching result, analyzing the stable sensor signal data and the target human body features through a stride length recognition model to generate a stride length estimation result. Thus, by performing noise reduction processing on the sensor data and then performing stride length recognition modal matching, stride length recognition is performed according to the modal matching result, combined with the sensor data and human body features, to achieve stride length recognition under different modalities, thereby achieving the technical effect of improving the efficiency and accuracy of stride length recognition.

[0060] Example 2

[0061] Based on the same inventive concept as the hybrid sensor positioning method for pedestrian stride length estimation in the aforementioned embodiment, this application also provides a hybrid sensor positioning system for pedestrian stride length estimation, please refer to the attached Figure 2 , the system comprising:

[0062] A connection establishing unit 11, the connection establishing unit 11 is used to establish a communication connection with an intelligent sensor module, wherein the intelligent sensor module is embedded in an intelligent terminal;

[0063] A human body feature collection unit 12 is used to collect human body features of a target pedestrian to obtain target human body features;

[0064] A signal noise reduction unit 13 is configured to receive the sensor signal data collected by the intelligent sensor module and input the signal into a noise reduction processing network to perform signal noise reduction and generate stable sensor signal data;

[0065] A modal matching unit 14 is configured to perform bipedal balance state recognition based on the stable sensor signal data, and perform stride length recognition modal matching according to the balance state recognition result to obtain a modal matching result, wherein the stride length recognition mode includes a single-step mode or a double-step mode;

[0066] The step length recognition unit 15 is used to analyze the stable sensing signal data and the target human body characteristics according to the modal matching result through a step length recognition model to generate a step length estimation result.

[0067] Furthermore, the system further comprises:

[0068] The intelligent sensing module at least includes an accelerometer, a gyroscope and a magnetometer, and the accelerometer, the gyroscope and the magnetometer are packaged and embedded in the intelligent terminal.

[0069] Furthermore, the system further comprises:

[0070] a noise adding unit, configured to add noise to the sensing signal data to obtain a damaged sensing signal;

[0071] A network construction unit, configured to construct the noise reduction processing network based on an autoencoder, wherein the noise reduction processing network includes an encoder and a decoder;

[0072] A deep feature extraction unit is used to perform deep feature extraction on the damage sensing signal using the noise reduction processing network, reconstruct the original signal, and obtain the stable sensing signal data.

[0073] Furthermore, the system further comprises:

[0074] a noise-added data acquiring unit, configured to acquire first noise-added data;

[0075] a noise fusion unit configured to obtain a historical sensor signal acquisition sample set and perform noise fusion on the historical sensor signal acquisition sample set with the first noise-added data to obtain a first damage sensor signal sample set;

[0076] A reconstruction training unit is used to construct a signal reconstruction training function, and train the encoder and the decoder in combination with the first damaged sensor signal sample set and the historical sensor signal acquisition samples to generate the noise reduction processing network.

[0077] Furthermore, the system further comprises:

[0078] A function construction unit, wherein the function construction unit is used to construct the signal reconstruction training function:

[0079]

[0080] Among them, L(x,y) is the signal reconstruction loss value, x i is the i-th damage sensing signal sample in the first damage sensing signal sample set, y i is a reconstructed sensing signal output after the encoder and the decoder reconstruct the original signal of the i-th damage sensing signal sample, and N is the total number of samples in the first damage sensing signal sample set.

[0081] Furthermore, the system further comprises:

[0082] a peak-valley identification unit, configured to perform peak-valley identification on the stable sensor signal data to obtain peak-valley time series data;

[0083] A single-step cutting unit, configured to perform single-step cutting on the peak and trough time series data based on single-step periodic features to obtain a continuous single-step feature sequence;

[0084] a fluctuation analysis unit, configured to perform single-step duration fluctuation analysis on the continuous single-step feature sequence under continuous single-step cycles to obtain a double-foot balance state index;

[0085] A modal matching result acquisition unit is configured to determine that if the double-foot balance state index is greater than or equal to a predetermined balance index, the modal matching result is a two-step mode; and if the double-foot balance state index is less than the predetermined balance index, the modal matching result is a single-step mode.

[0086] Furthermore, the system further comprises:

[0087] A sample acquisition unit, the sample acquisition unit is used to acquire walking test sensor data samples and walking state time series samples;

[0088] a step node positioning unit, the step node positioning unit being configured to position a step node sequence based on the walking state time series sample;

[0089] a data segmentation unit, configured to perform data segmentation on the walking test sensor data sample based on the step number node sequence to obtain a plurality of single-step cycle test data;

[0090] The single-step cycle feature acquisition unit is used to perform peak and trough feature extraction on the multiple single-step cycle test data and perform concentration value evaluation to obtain the single-step cycle feature.

[0091] Furthermore, the system further comprises:

[0092] a period superposition unit configured to perform period superposition on the single-step periodic feature to obtain a double-step periodic feature if the mode matching result is a double-step mode;

[0093] a step length estimation result acquisition unit, configured to extract periodic data from the stable sensing signal data based on the double-step periodic feature, input the double-step sensing periodic data and the target human body feature into the step length recognition model for analysis, and generate the step length estimation result;

[0094] The step length estimation result carries identification information of the modality matching result.

[0095] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. Figure 1 The hybrid sensor positioning method and specific examples for pedestrian stride length estimation in Example 1 are also applicable to the hybrid sensor positioning system for pedestrian stride length estimation in this embodiment. Through the detailed description of the hybrid sensor positioning method for pedestrian stride length estimation above, those skilled in the art can clearly understand the hybrid sensor positioning system for pedestrian stride length estimation in this embodiment. Therefore, for the sake of brevity, it will not be described in detail here. As for the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple. For relevant details, please refer to the method description.

[0096] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present application. Various modifications to these embodiments will be readily apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application is not limited to the embodiments shown herein, but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.

[0097] Obviously, those skilled in the art may make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application is intended to include these modifications and variations.

Claims

1. A hybrid sensor positioning system for pedestrian step length estimation, characterized in that: include: A connection establishing unit, the connection establishing unit is used to establish a communication connection with an intelligent sensor module, wherein the intelligent sensor module is embedded in an intelligent terminal; A human body feature collection unit, the human body feature collection unit is used to collect human body features of a target pedestrian to obtain target human body features; A signal noise reduction unit, which is used to receive the sensor signal data collected by the intelligent sensor module and input it into a noise reduction processing network to perform signal noise reduction and generate stable sensor signal data; A mode matching unit, the mode matching unit is used to perform double-foot balance state recognition based on the stable sensor signal data, and perform step length recognition mode matching according to the balance state recognition result to obtain a mode matching result, wherein the step length recognition mode includes a single-step mode or a double-step mode; The step size identification modal matching is performed according to the equilibrium state identification result to obtain the modal matching result, including: A peak-valley identification unit, the peak-valley identification unit is used to identify the peaks and troughs of the stable sensor signal data to obtain peak-valley time series data; A single-step cutting unit, the single-step cutting unit is used to perform single-step cutting on the wave crest and wave trough time series data based on the single-step period feature to obtain a continuous single-step feature sequence; A fluctuation analysis unit, the fluctuation analysis unit is used to perform single-step duration fluctuation analysis under continuous single-step cycles on the continuous single-step feature sequence to obtain a double-foot balance state index; A modal matching result acquisition unit, wherein if the double-foot balance state index is greater than or equal to a predetermined balance index, the modal matching result is a double-step mode, and if the double-foot balance state index is less than a predetermined balance index, the modal matching result is a single-step mode; A step length recognition unit is used to analyze the stable sensor signal data and the target human body characteristics according to the modal matching result through a step length recognition model to generate a step length estimation result.

2. The hybrid sensor positioning system for pedestrian step length estimation according to claim 1, characterized in that: The intelligent sensing module at least includes an accelerometer, a gyroscope and a magnetometer, and the accelerometer, the gyroscope and the magnetometer are packaged and embedded in the intelligent terminal.

3. The hybrid sensor positioning system for pedestrian step length estimation according to claim 1, characterized in that: Also includes: A noise adding unit, the noise adding unit is used to add noise to the sensor signal data to obtain a damaged sensor signal; A network construction unit, the network construction unit is used to construct the noise reduction processing network based on the autoencoder, wherein the noise reduction processing network includes an encoder and a decoder; A deep feature extraction unit is used to perform deep feature extraction on the damage sensor signal using the noise reduction processing network, reconstruct the original signal, and obtain the stable sensor signal data.

4. The hybrid sensor positioning system for pedestrian step length estimation as claimed in claim 3, characterized in that: Also includes: a noise adding data acquiring unit, the noise adding data acquiring unit being used to acquire first noise adding data; A noise fusion unit, the noise fusion unit is used to obtain a historical sensor signal acquisition sample set, and perform noise fusion with the first noise addition data to obtain a first damage sensor signal sample set; A reconstruction training unit is used to construct a signal reconstruction training function, and train the encoder and the decoder in combination with the first damaged sensor signal sample set and the historical sensor signal acquisition samples to generate the noise reduction processing network.

5. The hybrid sensor positioning system for pedestrian step length estimation as claimed in claim 4, characterized in that: Also includes: A function construction unit, wherein the function construction unit is used to construct the signal reconstruction training function: ， in, Reconstruct the loss value for the signal, is the first damage sensing signal sample set i damage sensing signal samples, To process the first i A reconstructed sensing signal is output after reconstructing the original signal using damage sensing signal samples, and N is the total number of samples in the first damage sensing signal sample set.

6. The hybrid sensor positioning system for pedestrian step length estimation as claimed in claim 1, characterized in that: Also includes: A sample acquisition unit, the sample acquisition unit is used to acquire walking test sensor data samples and walking state time series samples; A step node positioning unit, the step node positioning unit is used to locate the step node sequence based on the walking state time series sample; A data segmentation unit, the data segmentation unit is used to perform data segmentation on the walking test sensor data sample based on the step number node sequence to obtain a plurality of single-step cycle test data; A single-step cycle feature acquisition unit is used to perform peak and trough feature extraction on the multiple single-step cycle test data, and perform concentration value evaluation to obtain the single-step cycle feature.

7. The hybrid sensor positioning system for pedestrian step length estimation as claimed in claim 1, characterized in that: Also includes: A period superposition unit, wherein if the mode matching result is a double-step mode, the period superposition unit is used to perform period superposition on the single-step period feature to obtain a double-step period feature; a step length estimation result acquisition unit, the step length estimation result acquisition unit being used to perform period data extraction on the stable sensing signal data based on the double-step period feature, input the double-step sensing period data and the target human body feature into the step length recognition model for analysis, and generate the step length estimation result; Among them, the step size inference result carries identification information including the modality matching result.

8. A hybrid sensor positioning method for pedestrian step length estimation, characterized in that: The method comprises: Establishing a communication connection with an intelligent sensor module, wherein the intelligent sensor module is embedded in an intelligent terminal; Collect human features of the target pedestrian to obtain the target human features; Receive the sensor signal data collected by the intelligent sensor module and input it into the noise reduction processing network to perform signal noise reduction to generate stable sensor signal data; Performing double-foot balance state recognition based on the stable sensor signal data, performing step length recognition mode matching according to the balance state recognition result, and obtaining a mode matching result, wherein the step length recognition mode includes a single-step mode or a double-step mode; The step size identification modal matching is performed according to the equilibrium state identification result to obtain the modal matching result, including: A peak-valley identification unit, the peak-valley identification unit is used to identify the peaks and troughs of the stable sensor signal data to obtain peak-valley time series data; A single-step cutting unit, the single-step cutting unit is used to perform single-step cutting on the wave crest and wave trough time series data based on the single-step period feature to obtain a continuous single-step feature sequence; A fluctuation analysis unit, the fluctuation analysis unit is used to perform single-step duration fluctuation analysis under continuous single-step cycles on the continuous single-step feature sequence to obtain a double-foot balance state index; A modal matching result acquisition unit, wherein if the double-foot balance state index is greater than or equal to a predetermined balance index, the modal matching result is a double-step mode, and if the double-foot balance state index is less than a predetermined balance index, the modal matching result is a single-step mode; According to the modal matching result, the stable sensing signal data and the target human body characteristics are analyzed through a step length recognition model to generate a step length estimation result.

Images