Construction access control method and system based on voice recognition

Abstract

The application discloses a kind of based on voice recognition's building construction access control method and system, it is related to voice recognition technical field.The method includes: in the case where target user enters identity verification area, verification order structure is obtained, verification order structure is used to show the priority order between multiple identity verification methods;In the case where the first order of verification order structure is voice recognition method, the first audio to be verified that target user inputs on site, the reference audio of target user and the environmental audio of identity verification area are obtained in advance;Based on environmental audio, the first audio to be verified is de-noised, and the second audio to be verified is obtained;Second audio to be verified and reference audio are compared, and the voice verification result of voice recognition method is obtained, to make the access of target user based on voice verification result control.This application can improve the efficiency and accuracy of access control.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of access control, in particular to a construction site access control method and system based on voice recognition. BACKGROUND

[0002] The construction site has a large flow of personnel, various roles, a complex construction environment, and high-risk operation links such as high-altitude operation and large-scale mechanical operation. If unauthorized personnel enter at will, not only will the construction process be disrupted, but also safety accidents will be easily caused, endangering the safety of personnel life and property. Therefore, establishing a reliable and efficient personnel access control system and strictly controlling the access personnel are key links in construction management.

[0003] Currently, two recognition methods, face recognition and access card recognition, are commonly used in construction to control personnel access. Face recognition verifies identity based on facial features, and a camera captures images and compares them with pre-stored face templates to determine legitimacy. Access card recognition uses cards as carriers, stores personnel information in the cards, and matches the read information with the system database to determine access permissions. Some construction sites support both recognition methods to improve the convenience of access control.

[0004] However, the construction site requires personnel to wear protective equipment such as safety helmets and masks, which can seriously interfere with face recognition, block key facial features, and make it difficult for the camera to clearly capture information, leading to recognition bias or even failure to recognize. In addition, the use of access cards is not convenient, as workers hold tools in both hands while working, which can interrupt work and reduce efficiency, and the cards are easily lost, damaged, or forgotten. Furthermore, when multiple recognition methods are supported, there is a lack of scientific judgment mechanism, affecting the accuracy of personnel access control. As a result, the efficiency and accuracy of access control are poor. SUMMARY

[0005] The embodiments of the present application provide a construction site access control method and system based on voice recognition, which can improve the efficiency and accuracy of access control.

[0006] In a first aspect, the embodiments of the present application provide a construction site access control method based on voice recognition, comprising:

[0007] In the case that the target user enters the identity verification area, a verification sequence structure is obtained, which is used to represent the priority order between multiple identity verification methods;

[0008] In the case that the first order of the verification sequence structure is the voice recognition method, the first to-be-verified audio input by the target user on site, the reference audio of the target user pre-stored, and the environmental audio of the identity verification area are obtained;

[0009] Based on the environmental audio, the first to be verified audio is denoised to obtain the second to be verified audio;

[0010] The second to be verified audio is compared with the reference audio to obtain a voice verification result in a voice recognition mode, so as to control the access of the target user based on the voice verification result.

[0011] Further, the present application also proposes that, in the case that the target user enters the identity verification area, before the verification sequence structure is obtained, the method further comprises:

[0012] Obtaining historical to-be-verified information of various identity verification modes and attendance record information in a historical time period;

[0013] Based on the historical to-be-verified information and the corresponding attendance record information, the verification effectiveness of each identity verification mode in each attendance record information is determined respectively;

[0014] Based on the verification effectiveness of each identity verification mode in each attendance record information and the verification order of each identity verification mode in each attendance record information, the verification preference degree of each identity verification mode in each verification order is determined respectively;

[0015] Based on the verification preference degree of each identity verification mode in each verification order, the verification sequence structure is constructed.

[0016] Further, the present application also proposes that, based on the historical to-be-verified information and the corresponding attendance record information, the verification effectiveness of each identity verification mode in each attendance record information is determined respectively, comprising:

[0017] Based on the difference between each historical to-be-verified information and the corresponding reference verification information, the verification possibility index of each historical to-be-verified information is determined respectively;

[0018] Based on each verification possibility index included in each identity verification mode, the verification reliability of each identity verification mode is determined respectively;

[0019] Based on each attendance record information and the verification reliability of each identity verification mode, the verification effectiveness of each identity verification mode in each attendance record information is determined respectively.

[0020] Further, the present application also proposes that the historical to-be-verified information is historical to-be-verified audio, and the reference verification information is reference audio;

[0021] Based on the difference between each historical to-be-verified information and the corresponding reference verification information, the verification possibility index of each historical to-be-verified information is determined respectively, comprising:

[0022] The audio segment in the historical to-be-verified audio which is consistent with the waveform of the reference audio is determined as an effective audio segment;

[0023] performing short-time Fourier transform on the effective audio segment to obtain instantaneous phase difference and amplitude spectrum of the effective audio segment in each time frame;

[0024] connecting harmonic peak points in adjacent amplitude spectrum in sequence to form a spectrum envelope line;

[0025] selecting a maximum envelope line slope difference from envelope line slope differences, the envelope line slope difference being a slope difference between adjacent spectrum envelope lines;

[0026] determining a verification possibility index of the historical audio to be verified by using variance of the instantaneous phase difference, the maximum envelope line slope difference and a spectrum centroid mean value, the spectrum centroid mean value being a mean value of spectrum centroids corresponding to each time frame.

[0027] Further, the present application also proposes that, based on the verification possibility indexes included in each identity verification mode, the verification reliability of each identity verification mode is determined respectively, comprising:

[0028] obtaining information accuracy of the first historical audio to be verified information, the first historical audio to be verified information being historical audio to be verified information corresponding to a target identity verification mode, the target identity verification mode being any one identity verification mode, the information accuracy being a matching degree between the first historical audio to be verified information and reference verification information;

[0029] performing mean value processing on the verification possibility index and the information accuracy of each first historical audio to be verified information respectively to obtain a verification possibility index mean value and an information accuracy mean value;

[0030] determining local verification reliability of each first historical audio to be verified information by using the verification possibility index, the information accuracy, the verification possibility index mean value and the information accuracy mean value of each first historical audio to be verified information;

[0031] determining target verification reliability of the target identity verification mode based on the local verification reliability of each first historical audio to be verified information.

[0032] Further, the present application also proposes that, based on the verification reliability of each identity verification mode and each attendance record information, the verification effectiveness of each identity verification mode in each attendance record information is determined respectively, comprising:

[0033] obtaining a first attendance record vector of the first attendance record information and a first key feature vector in the second historical audio to be verified information, the second historical audio to be verified information being historical audio to be verified information of a target identity verification mode corresponding to the first attendance record information, the target identity verification mode being any one identity verification mode;

[0034] obtain a second attendance record vector of second attendance record information and a second key feature vector in third historical to-be-verified information, the second attendance record information is a previous attendance record information of the first attendance record information, and the third historical to-be-verified information is historical to-be-verified information of the target identity verification mode corresponding to the second attendance record information;

[0035] determine the verification effectiveness of the target identity verification mode in the first attendance record information by using the first attendance record vector, the first key feature vector, the second attendance record vector, the second key feature vector and the verification reliability of the target identity verification mode.

[0036] Further, the present application also proposes that, based on the verification effectiveness of each identity verification mode in each attendance record information and the verification order of each identity verification mode in each attendance record information, the verification preference degree of each identity verification mode in each verification order is determined respectively, comprising:

[0037] select third attendance record information from each attendance record information, and the third attendance record information is used to represent the attendance record information of the target identity verification mode in the target verification order;

[0038] determine the verification preference degree of the target identity verification mode in the target verification order by using the verification effectiveness of the target identity verification mode in each third attendance record information and the number of the third attendance record information.

[0039] Further, the present application also proposes that the second to-be-verified audio is compared with the reference audio to obtain the voice verification result of the voice recognition mode, so as to control the access of the target user based on the voice verification result, comprising:

[0040] compare the second to-be-verified audio with the reference audio to obtain the voice verification result of the voice recognition mode;

[0041] in the case that the voice verification result indicates that the identity verification is passed, send an open door instruction signal to the controller, so that the controller controls the opening of the gate based on the open door instruction signal, so as to control the access of the target user;

[0042] in the case that the voice verification result indicates that the identity verification is not passed, perform auxiliary identity verification based on the verification sequence structure, so as to control the access of the target user based on the auxiliary identity verification result.

[0043] Further, the present application also proposes that the auxiliary identity verification is performed based on the verification sequence structure, so as to control the access of the target user based on the auxiliary identity verification result, comprising:

[0044] perform auxiliary identity verification based on the identity verification mode in the next order of the verification sequence structure to obtain an auxiliary identity verification result;

[0045] In the case that the auxiliary authentication result indicates that the authentication fails, a cycle is returned to perform auxiliary authentication based on the next order of the authentication order structure, to obtain an auxiliary authentication result, until a preset stop condition is reached, the preset stop condition being that the auxiliary authentication result indicates that the authentication passes or each authentication mode of the authentication order structure is traversed;

[0046] In the case that the auxiliary authentication result indicates that the authentication passes, an artificial verification process is triggered, and the access of the target user is controlled based on an artificial verification result.

[0047] In a second aspect, the application provides a building construction access control system based on voice recognition, comprising:

[0048] The structure acquisition module is configured to acquire an authentication order structure in the case that the target user enters the authentication area, the authentication order structure being used to represent a priority order among the multiple authentication modes;

[0049] The audio acquisition module is configured to acquire, in the case that the first order of the authentication order structure is the voice recognition mode, a first to-be-verified audio input by the target user on site, a reference audio of the target user stored in advance, and an environmental audio of the authentication area.

[0050] The audio processing module is configured to perform denoising processing on the first to-be-verified audio based on the environmental audio, to obtain a second to-be-verified audio.

[0051] The result determination module is configured to compare the second to-be-verified audio with the reference audio, to obtain a voice authentication result of the voice recognition mode, so as to control the access of the target user based on the voice authentication result.

[0052] The application has the following beneficial effects:

[0053] The building construction access control method based on voice recognition provided by the embodiment of the present application comprises the following steps: setting a verification sequence structure; when the first order is voice recognition, obtaining a first to-be-verified audio input by a target user on site, a reference audio stored in advance, and an environmental audio; performing denoising processing on the first to-be-verified audio based on the environmental audio to obtain a second to-be-verified audio; and comparing the second to-be-verified audio with the reference audio to obtain a voice verification result. The voice recognition is not affected by the shielding of the protection device, and the worker does not need to put down the tool in hand to perform additional operations, thereby avoiding the situation that the card is taken to interrupt the work, and improving the access efficiency. Meanwhile, the denoising processing can effectively reduce the environmental noise interference, so that the second to-be-verified audio is clearer, the voice verification result obtained by comparing the second to-be-verified audio with the reference audio is more accurate, and in combination with the verification sequence structure, the multiple recognition modes are used scientifically and reasonably, the deficiencies of the traditional method in the use of multiple recognition modes are made up, and the accuracy of the personnel access control is comprehensively improved. Therefore, the present application can effectively solve the problems of the traditional recognition mode in the building construction scene, and can improve the efficiency and accuracy of the access control. BRIEF DESCRIPTION OF DRAWINGS

[0054] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, and the advantages thereof, the drawings needed to be used in the embodiments or the prior art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor based on these drawings.

[0055] Figure 1 The flowchart of the first building construction access control method based on voice recognition provided by an embodiment of the present application is shown in the figure.

[0056] Figure 2 The flowchart of the second building construction access control method based on voice recognition provided by an embodiment of the present application is shown in the figure.

[0057] Figure 3 The flowchart of S202 provided by an embodiment of the present application is shown in the figure.

[0058] Figure 4 The flowchart of S104 provided by an embodiment of the present application is shown in the figure.

[0059] Figure 5 The structural diagram of the building construction access control system based on voice recognition provided by an embodiment of the present application is shown in the figure. DETAILED DESCRIPTION

[0060] In order to further illustrate the technical means and effects taken by the present application to achieve the predetermined object of the application, the specific implementation, structure, features and effects of the speech recognition-based building construction access control method and system according to the present application are described in detail as follows in combination with the drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures or characteristics in one or more embodiments can be combined in any suitable form.

[0061] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present application belongs.

[0062] It should be noted that the acquisition, storage, use, processing, etc. of data in the technical solution of the present application comply with the relevant provisions of laws and regulations.

[0063] It should be noted that in the embodiments of the present application, some industry existing solutions of software, components, models, etc. may be mentioned, which should be considered as exemplary, and the purpose is only to illustrate the feasibility in the implementation of the technical solution of the present application, but it does not mean that the applicant has or will necessarily use the solution.

[0064] In the traditional existing building construction access control system, the identity verification mode relying on single biometric feature or physical carrier has significant limitations. The face recognition technology is easily affected by the shielding of safety protection equipment, resulting in a decrease in feature extraction accuracy. The infrared light supplement module is difficult to accurately reconstruct the three-dimensional face model under strong environmental light interference. The access card verification relies on physical contact type interaction, which is easy to cause radio frequency signal collision in the multi-person concurrent scene, and cannot verify the biological correlation between the card holder and the identity information in real time. Although the multi-modal verification system adopts a parallel verification mechanism, it lacks a dynamic priority scheduling algorithm. The data collection timing conflict of different verification modes will cause system resource competition, resulting in a nonlinear increase in the overall verification response time.

[0065] When facing the above problems, the present application first considers how to break through the dependence of traditional verification methods on physical carriers and facial features. Noise interference in the complex acoustic environment of the construction site directly affects the accuracy of speech recognition, and a targeted noise reduction mechanism needs to be designed. The dynamic scheduling problem of the multi-modal verification system requires the establishment of a scientific verification sequence decision model to avoid system congestion caused by resource competition. The low efficiency of traditional verification methods in concurrent scenes prompts the introduction of a priority strategy, and the non-contact nature of voice interaction can alleviate the operational burden when people hold objects. The present application further analyzes that different time periods and environmental factors have a dynamic impact on the effectiveness of the verification method, and a verification sequence structure that can be adaptively adjusted needs to be constructed, in which the generation of the verification sequence needs to combine the reliability features in the historical verification data.

[0066] To this end, as shown in Figure 1 The application provides a construction access control method based on voice recognition, which can be applied to a server and comprises the following steps S101-S104:

[0067] S101, in the case that a target user enters an identity verification area, a verification sequence structure is obtained, which is used to represent the priority order among multiple identity verification methods;

[0068] S102, in the case that the first order of the verification sequence structure is a voice recognition method, a first to-be-verified audio input by the target user on site, reference audio of the target user pre-stored, and environmental audio of the identity verification area are obtained;

[0069] S103, based on the environmental audio, the first to-be-verified audio is denoised to obtain second to-be-verified audio;

[0070] S104, the second to-be-verified audio is compared with the reference audio to obtain a voice verification result of the voice recognition method, so as to control the access of the target user based on the voice verification result.

[0071] In this embodiment, the verification sequence structure refers to the logical arrangement structure of the priority order among multiple identity verification methods, which can be dynamically adjusted or pre-set to achieve the execution order of the identity verification process according to different scene requirements, thereby improving the verification efficiency.

[0072] The identity verification area refers to a pre-defined spatial range for performing an identity verification operation, which can be realized by infrared induction or camera positioning technology, and its function is to ensure that the triggering time of the identity verification process matches the actual position of the user through physical or virtual boundary limitation.

[0073] The environmental audio refers to the background sound signal existing in the identity verification area, which can be realized by directional microphone array collection, and its function is to provide noise samples for subsequent audio denoising and eliminate the negative impact of environmental interference on voice recognition.

[0074] The denoising process refers to noise reduction and enhancement of the target voice signal based on environmental noise characteristics, which can be realized by spectral subtraction or adaptive filtering algorithm, and its function is to improve the subsequent voiceprint comparison accuracy by eliminating the spectral overlap of environmental noise and voice signal.

[0075] The reference audio refers to the pre-registered user voiceprint feature data, which can be realized by mel-frequency cepstral coefficient modeling, and its function is to provide a comparison benchmark for real-time collected voice and ensure the uniqueness of the biological characteristics of identity verification.

[0076] The core invention point of the present application is that by combining dynamic priority verification sequence with multi-modal identity verification technology, an adaptive speech recognition framework is constructed in a complex construction scene, thereby improving the accuracy and execution efficiency of personnel access control under the premise of ensuring security.

[0077] As an example, an identity verification area is set at the entrance of a construction site, equipped with a high-sensitivity microphone array and an environmental noise collection device. When the target user enters the identity verification area, the server first obtains a verification sequence structure. The verification sequence structure represents the priority order of multiple identity verification methods, which is used to dynamically adjust the verification strategy.

[0078] When the first order of the verification sequence structure is the speech recognition method, the server obtains three types of audio data: the first to-be-verified audio input by the target user on site, the pre-stored reference audio of the target user, and the environmental audio of the identity verification area. Specifically, the target user speaks a preset password word into the microphone, and the server records it as the first to-be-verified audio.

[0079] Next, the server uses the environmental audio to denoise the first to-be-verified audio to obtain the second to-be-verified audio. This step improves the accuracy of speech recognition by eliminating environmental noise. Specifically, the server can use an adaptive filtering algorithm to denoise the first to-be-verified audio based on the collected environmental noise to obtain the second to-be-verified audio.

[0080] Then, the processed second to-be-verified audio is compared with the reference audio to obtain a speech verification result. The access permission of the target user is controlled based on the result. Specifically, the dynamic time warping algorithm can be used to compare the second to-be-verified audio with the reference audio to calculate the similarity score, and then the speech verification result is obtained according to the similarity score. Then, if the speech verification result indicates that the identity verification is passed, an opening signal is sent to the gate to allow the target user to enter. If the speech verification result indicates that the identity verification is not passed, the next verification method such as face recognition or access card verification will be switched according to the verification sequence structure.

[0081] The introduction of the verification sequence structure enables the server to flexibly select the most suitable verification method according to different scenarios. The collection and use of environmental audio effectively solve the problem of interference of complex acoustic environment of construction sites on speech recognition. Denoising further improves the quality of speech samples. By comparing the processed second to-be-verified audio with the reference audio, the server can reliably determine the user's identity, thereby achieving precise access control.

[0082] By the embodiment, the verification sequence structure is first set, when the first order is voice recognition, the first to-be-verified audio input by the target user on site, the reference audio and the environmental audio are acquired, the first to-be-verified audio is denoised based on the environmental audio to obtain the second to-be-verified audio, and the second to-be-verified audio is compared with the reference audio to obtain a voice verification result. The voice recognition is not affected by the shielding of the protection device, the worker does not need to put down the tool in hand to perform additional operations, the situation that the card is taken to interrupt the work is avoided, and the efficiency of the entry and exit is improved. Meanwhile, the denoising processing can effectively reduce the environmental noise interference, the second to-be-verified audio is clearer, the voice verification result obtained by comparing the second to-be-verified audio with the reference audio is more accurate, and in combination with the verification sequence structure, various recognition modes are scientifically and reasonably used, the deficiency of the traditional method in the use of various recognition modes is made up, and the accuracy of the personnel entry and exit control is comprehensively improved. Therefore, the present application can effectively solve the problems of the traditional recognition mode in the construction scene, and can improve the efficiency and accuracy of the entry and exit control.

[0083] In some schemes of the present application, the priority of the identity verification mode is sorted by using a fixed rule when the verification sequence structure is constructed, and the adaptability to the dynamic changes in the actual scene is lacking. The fixed sorting method cannot reflect the effectiveness difference of the verification modes in different time periods, the verification sequence cannot be dynamically adjusted according to the actual verification effect, and the accuracy and efficiency of the entry and exit control are affected.

[0084] To this end, as shown in Figure 2 The present application further provides that before S101, the construction entry and exit control method based on voice recognition can further include the following S201 to S204:

[0085] S201, historical to-be-verified information of various identity verification modes and attendance record information in a historical time period are acquired;

[0086] S202, the verification effectiveness of each identity verification mode in each attendance record information is determined based on each historical to-be-verified information and the corresponding attendance record information;

[0087] S203, the verification preferred degree of each identity verification mode in each verification order is determined based on the verification effectiveness of each identity verification mode in each attendance record information and the verification order of each identity verification mode in each attendance record information;

[0088] S204, the verification sequence structure is constructed based on the verification preferred degree of each identity verification mode in each verification order.

[0089] In the embodiment, the historical to-be-verified information contains actual verification data of the identity verification manners in different time periods, and the attendance record information reflects actual access records corresponding to the verification operations. The verification validity is calculated through the association of the historical to-be-verified information and the attendance record, and is used to evaluate the reliability of the verification manners in a specific time period. The verification order refers to the priority position of the verification manners in the historical verification process, and the verification preference is calculated by integrating the verification validity and the historical verification order. The verification sequence structure is generated by integrating the preferences of the verification manners in different orders.

[0090] Specifically, the historical time period is divided into multiple continuous intervals, and the verification request data and the actual attendance results corresponding to each identity verification manner are collected in each interval. For example, 100 face recognition verification requests are collected in a time period, 85 of which are verified and recorded as valid attendance, 80 access card verification requests are collected, and 76 of which are verified and recorded as valid attendance. The verification validity is determined by calculating the ratio of the number of verification passes to the matching degree of the attendance record. The verification validity of a verification manner in multiple attendance record information is averaged to serve as a comprehensive effectiveness index of the manner.

[0091] The verification order is determined by counting the starting order of each manner in the historical verification process. For example, 50 voice recognition verifications are started preferentially in a time period, 40 of which are verified and the subsequent verification steps are skipped. The verification preference is calculated by weighting the verification validity and the order frequency. For example, the historical frequency of voice recognition in the second order is 30%, and the verification validity is 0.92, so the verification preference is 0.92×0.3=0.276. The verification preferences of each verification manner in different orders are compared horizontally, and the verification manner with the highest verification preference is assigned to the corresponding verification order to form a dynamically adjusted verification sequence structure.

[0092] As an example, the historical to-be-verified information and the attendance record information of various identity verification manners in a historical time period are first obtained. For example, the historical voice data, face images and fingerprint images of voice recognition, face recognition and fingerprint recognition in the past month are obtained, as well as the corresponding attendance clock-in records.

[0093] Based on the historical to-be-verified information and the corresponding attendance record information, the verification validity of each identity verification manner in each attendance record information is determined. Specifically, the matching degree of the historical to-be-verified information and the pre-stored reference information can be compared, and the accuracy of the attendance record can be combined to calculate the verification effectiveness score of each identity verification manner in each attendance record.

[0094] Further, based on the verification effectiveness of each authentication mode in each attendance record information and the verification order of each authentication mode in each attendance record information, the verification preference degree of each authentication mode in each verification order is determined respectively. For example, the verification effectiveness score distribution of each authentication mode in different verification orders is counted, and the verification preference degree of each authentication mode in each verification order is calculated.

[0095] Finally, based on the verification preference degree of each authentication mode in each verification order, the verification order structure is constructed. Further, according to the high and low of the verification preference degree, each authentication mode can be sorted to generate a priority order list as the final verification order structure.

[0096] Through the embodiment, the identity authentication order can be dynamically optimized according to historical verification data, and the verification efficiency and accuracy are improved. Therefore, the most suitable combination of authentication modes can be selected to adapt to different working environments and personnel characteristics. Further, the scheme reduces unnecessary verification attempts, shortens the personnel access time, and improves the management efficiency of the construction site.

[0097] In some schemes of the present application, the difference between the historical to-be-verified information and the reference verification information is not fully quantified, which leads to the lack of objective basis for determining the verification possibility index, affects the accuracy of subsequent verification reliability calculation, and further affects the determination of verification effectiveness.

[0098] To this end, as shown in Figure 3 the present application further provides that S202 can specifically include the following S301 to S303:

[0099] S301, based on the difference between each historical to-be-verified information and the corresponding reference verification information, the verification possibility index of each historical to-be-verified information is determined respectively;

[0100] S302, based on each verification possibility index included in each authentication mode, the verification reliability of each authentication mode is determined respectively;

[0101] S303, based on each attendance record information and the verification reliability of each authentication mode, the verification effectiveness of each authentication mode in each attendance record information is determined respectively.

[0102] In the embodiment, the historical to-be-verified information is various information used for verifying the identity of employees in past attendance scenarios. For example, it can be the password input by the employee, image data of face recognition, fingerprint data, etc. These information are original data received by the server for identity authentication.

[0103] The attendance record information is information recording the attendance of employees, usually including employee identification (such as employee number, name, etc.), attendance time (entering and leaving punch-in time, etc.), which is a data record reflecting the actual attendance behavior of employees.

[0104] The reference verification information is standard information pre-stored in the system, used for comparison with historical verification information to determine whether the employee's identity is legal. For example, for password verification, the reference verification information is the correct password set by the employee in advance; for face recognition verification, the reference verification information is the employee's facial feature data collected and stored in advance.

[0105] The verification possibility index is a quantitative index for measuring the matching degree of historical verification information and reference verification information, which reflects the possibility of historical verification information passing identity verification. For example, in face recognition, the similarity between historical verification image and reference verification image can be calculated to determine the verification possibility index, the higher the similarity, the greater the verification possibility index.

[0106] The verification reliability is an index for measuring the ability of a certain identity verification method to accurately identify the identity of employees in the entire attendance system. It considers the verification possibility index of all historical verification information under this identity verification method, reflecting the stability and accuracy of the verification method in actual application. For example, if the verification possibility index of most historical verification information under a certain identity verification method is high, the verification reliability of the verification method is high.

[0107] The verification effectiveness is an index for measuring the effectiveness of a certain identity verification method in a specific attendance record scenario. It combines the characteristics of the attendance record information itself and the verification reliability of the identity verification method, reflecting the applicability and accuracy of the verification method in a specific attendance record. For example, in a certain attendance record information, if the verification possibility index of the employee corresponding to the attendance record information is generally high when using a certain identity verification method, and the verification reliability of the identity verification method is also high, the verification effectiveness of the identity verification method in the attendance record is high.

[0108] As an example, the verification possibility indicators of the historical to-be-verified information are determined based on the differences between the historical to-be-verified information and the corresponding reference verification information. For example, in the password verification mode, the historical to-be-verified information (i.e., the historical password input by the employee) is compared with the corresponding reference verification information (i.e., the correct password set in advance) character by character. If they are completely matched, the verification possibility indicator is set to the highest value. If there are some characters that are not matched, the verification possibility indicator can be calculated according to the number and position of the unmatched characters, for example, a certain score is subtracted from the highest value for each unmatched character, and finally a specific numerical value is obtained as the verification possibility indicator. In the fingerprint verification mode, the fingerprint feature points are extracted from the historical to-be-verified fingerprint data (i.e., the data collected when the employee historically punches the fingerprint) and the reference verification fingerprint data (i.e., the employee fingerprint feature data collected and stored in advance). The verification possibility indicator is determined by comparing the matching degree of the two sets of fingerprint feature points. For example, the proportion of the number of matched feature points to the total number of feature points can be calculated, and then the proportion is multiplied by 100 to obtain the verification possibility indicator.

[0109] Then, the verification reliabilities of the identity verification modes are determined based on the verification possibility indicators included in each identity verification mode. Specifically, the statistical quantities such as the average value and variance of the verification possibility indicators can be calculated. The average value reflects the average level of the matching degree between the historical to-be-verified information and the reference verification information in the identity verification mode, and the variance reflects the fluctuation of the matching degree. According to the statistical quantities such as the average value and variance, the verification reliability of the identity verification mode is determined in combination with a preset rule or model. For example, a threshold value can be set. If the average value is higher than the threshold value and the variance is small, it is considered that the verification reliability of the identity verification mode is high, and a higher reliability score can be given. If the average value is low or the variance is large, it is considered that the verification reliability is low, and a lower score is given.

[0110] Finally, the verification effectiveness of each identity verification mode in each attendance record information is determined based on the attendance record information and the verification reliabilities of the identity verification modes. Specifically, the verification effectiveness of the identity verification mode in the attendance record information is determined by a preset algorithm or model in combination with the verification reliability of the identity verification mode and the characteristics of the attendance record information. For example, if the attendance time in the attendance record information is in the peak period, and the verification accuracy of a certain identity verification mode is usually low in the peak period (obtained by historical data analysis), then even if the verification reliability of the identity verification mode is high, the verification effectiveness in the attendance record information will also be reduced accordingly.

[0111] Through the embodiment, the verification effectiveness of each authentication mode can be analyzed based on historical data, and the verification accuracy is improved. By considering multiple factors such as the difference between historical to-be-verified information and reference information, the relevance of attendance record information, and the like, the reliability of each authentication mode is comprehensively evaluated. This data-driven verification effectiveness analysis method can adapt to different construction environments and personnel characteristics, dynamically optimize the verification strategy, and improve the accuracy and efficiency of access control.

[0112] In some schemes of the present application, the determination method of the verification possibility index has the problem of insufficient accuracy in complex construction environments. Due to environmental noise interference, the difference between historical to-be-verified audio and reference audio is difficult to accurately quantify, resulting in deviation in subsequent verification reliability evaluation.

[0113] To this end, the present application further proposes that the historical to-be-verified information is historical to-be-verified audio, and the reference verification information is reference audio.

[0114] S301 can specifically include:

[0115] The audio segment in the historical to-be-verified audio that is consistent with the waveform of the reference audio is determined as an effective audio segment;

[0116] The effective audio segment is subjected to short-time Fourier transform processing to obtain the instantaneous phase difference and amplitude spectrum of the effective audio segment at each time frame;

[0117] The harmonic peak points in the adjacent amplitude spectra are sequentially connected to form a spectral envelope line;

[0118] The maximum envelope line slope difference is selected from the envelope line slope differences, and the envelope line slope difference is the slope difference between adjacent spectral envelope lines;

[0119] The verification possibility index of the historical to-be-verified audio is determined by using the variance of the instantaneous phase difference, the maximum envelope line slope difference, and the spectral centroid mean value, and the spectral centroid mean value is the mean value of the spectral centroids corresponding to each time frame.

[0120] In the embodiment, the effective audio segment is located by a waveform matching algorithm, ensuring that subsequent analysis focuses on the region with similar characteristics to the reference audio. The short-time Fourier transform performs time-frequency decomposition on the audio signal with a fixed time window, and the time window length is set to 20 milliseconds, and the adjacent window overlap rate is 50%. The spectral envelope line connects the harmonic peak points by linear interpolation, and the harmonic peak points are extracted in the amplitude spectrum by a peak detection algorithm. The maximum envelope line slope difference is obtained by calculating the absolute value difference of the slopes of adjacent envelope line segments, and the largest difference value is selected as the characteristic parameter. The spectral centroid mean value is obtained by calculating the arithmetic mean value of the spectral centroids of each time frame, and the spectral centroid is determined by the weighted sum of each frequency component and its amplitude.

[0121] Specifically, the extraction of the effective audio segment is achieved by a dynamic time warping algorithm, which aligns the time axis of the historical audio to be verified with the reference audio and marks the continuous region with waveform similarity exceeding the preset threshold. In the short-time Fourier transform process, a Hamming window is used to reduce spectral leakage, and the instantaneous phase difference of each time frame is obtained by comparing the phase information of adjacent frames. The construction of the spectral envelope further distinguishes noise interference from real speech features, as environmental noise usually exhibits wideband spectral characteristics, while the harmonic structure of the speech signal shows regular changes on the envelope. The maximum envelope slope difference is used to quantify the degree of mutation of the envelope shape, which can effectively reflect the transient component in the speech signal. The variance of the instantaneous phase difference describes the phase stability of the speech signal, and the mean of the spectral centroid represents the distribution of the signal energy in the frequency domain. These three types of parameters are fused by weighting to form the verification likelihood index. Thus, the determination process of the verification likelihood index integrates time domain, frequency domain and phase characteristics, improving the discrimination accuracy of speech verification in complex environments.

[0122] As an example, the verification likelihood index of the historical audio to be verified can be determined by the following formula 1:

[0123] Formula 1

[0124] In formula 1, is the verification likelihood index of the kth historical audio to be verified for the ith identity verification method, is the variance of the instantaneous phase difference in the kth historical audio to be verified, is the maximum envelope slope difference in the kth historical audio to be verified, is the mean of the spectral centroid in the kth historical audio to be verified.

[0125] Wherein, if the effective audio segment waveform of a historical audio to be verified and the reference audio is consistent, and the instantaneous phase difference of the effective audio segment fluctuates greatly, the slope of the frequency domain envelope fluctuates less (i.e. the energy distribution of the audio is more regular), and the spectral centroid is smaller, then the possibility of using the effective audio segment for correct identity verification is greater.

[0126] Through this embodiment, audio features can be analyzed from multiple dimensions, and more rich and stable audio feature parameters can be extracted. Thus, the accuracy and robustness of speech verification are improved, and the interference of environmental noise and other factors is reduced. Further, this scheme can adapt to the speech verification needs of different speakers and different environments, and improve the application effect of speech recognition in construction access control.

[0127] In some schemes of the above-mentioned embodiments of the present application, the verification effectiveness of the identity verification mode is determined by the historical to-be-verified information and the attendance record information, and then the verification sequence structure is constructed. However, in this process, the verification reliability is evaluated only based on the average of the verification possibility indexes, which is difficult to reflect the individual differences of different historical to-be-verified information, resulting in inaccurate evaluation of the verification reliability of the target identity verification mode.

[0128] To this end, the present application further proposes that S302 specifically can include:

[0129] obtaining information accuracy of first historical to-be-verified information, the first historical to-be-verified information being each historical to-be-verified information corresponding to a target identity verification mode, the target identity verification mode being any one identity verification mode, and the information accuracy being a matching degree between the first historical to-be-verified information and reference verification information;

[0130] respectively performing mean processing on the verification possibility indexes and the information accuracy of each first historical to-be-verified information to obtain a mean of the verification possibility indexes and a mean of the information accuracy;

[0131] determining local verification reliability of each first historical to-be-verified information by using the verification possibility indexes, the information accuracy, the mean of the verification possibility indexes and the mean of the information accuracy of each first historical to-be-verified information;

[0132] determining target verification reliability of the target identity verification mode based on the local verification reliability of each first historical to-be-verified information.

[0133] In this embodiment, the information accuracy is calculated by the matching degree between the first historical to-be-verified information and the reference verification information, for example, quantified by voiceprint similarity or text matching degree. The local verification reliability is determined by the deviation degree of the information accuracy and the verification possibility index, for example, obtained by multiplying the deviation degree and the information accuracy and then normalized. The target verification reliability is obtained by taking the maximum value or the mean of each local verification reliability.

[0134] Specifically, in determining the verification reliability of the target identity authentication manner, first, a plurality of historical to-be-verified information corresponding to the manner and corresponding reference verification information are extracted from the database. For each historical to-be-verified information, the matching degree thereof with the reference verification information is calculated to generate an information accuracy value between 0 and 1. At the same time, the mean value of the verification possibility indicators of the historical to-be-verified information is calculated to obtain a reference value. Subsequently, the verification possibility indicator of each historical to-be-verified information is compared with the reference value, the difference degree is calculated, and then multiplied by the information accuracy to obtain a local verification reliability reflecting the individual reliability. Finally, the maximum value of all local verification reliabilities is taken as the target verification reliability to avoid interference of abnormal data. For example, the information accuracy of a historical verification is 0.9, the verification possibility indicator is 0.85, and the mean value is 0.8. The difference degree is 0.05, and the local verification reliability is 0.9*0.05=0.045, which is normalized to 0.045 / 0.1=0.45. Through the method, the individual difference and the overall trend are comprehensively evaluated to improve the calculation accuracy of the target verification reliability.

[0135] As an example, the target verification reliability of the target identity authentication manner can be determined by the following formula 2:

[0136] Formula 2

[0137] In formula 2, is used to represent the verification reliability of the i-th identity authentication manner, is used to represent the verification possibility indicator of the k-th historical to-be-verified information of the i-th identity authentication manner, is used to represent the information accuracy of the k-th historical to-be-verified information of the i-th identity authentication manner, is used to represent the mean value of the verification possibility indicators of the i-th identity authentication manner, is used to represent the mean value of the information accuracy of the i-th identity authentication manner. is used to represent the number of historical to-be-verified information of the i-th identity authentication manner, and exp is used to represent the exponential function operation, represents an infinitesimal greater than 0, which is used to prevent the denominator from being 0.

[0138] wherein, when the value of the i-th identity authentication manner is smaller, the verification reliability is greater. That is, it means that the smaller the difference between the mean value of the verification possibility indicator and the information accuracy, the higher the verification reliability.

[0139] By evaluating the verification reliability of the identity verification mode based on the individual differences between different historical to-be-verified information, the evaluation accuracy of the verification reliability of the target identity verification mode can be improved, and thus the rationality of the constructed verification sequence structure can be improved.

[0140] In some schemes of the application, the verification validity evaluation is not accurate enough. Since the verification validity is calculated based only on the matching degree of the current attendance record and the historical to-be-verified information, the time sequence correlation between the attendance records is not considered, which cannot dynamically capture the actual effect changes of the verification mode in different time periods, and thus affects the optimization accuracy of the verification sequence structure.

[0141] To this end, the S303 can specifically include:

[0142] The first attendance record vector of the first attendance record information and the first key feature vector in the second historical to-be-verified information are obtained, the second historical to-be-verified information being the historical to-be-verified information of the target identity verification mode corresponding to the first attendance record information, and the target identity verification mode being any identity verification mode;

[0143] The second attendance record vector of the second attendance record information and the second key feature vector in the third historical to-be-verified information are obtained, the second attendance record information being the previous attendance record information of the first attendance record information, and the third historical to-be-verified information being the historical to-be-verified information of the target identity verification mode corresponding to the second attendance record information;

[0144] The verification validity of the target identity verification mode in the first attendance record information is determined by using the first attendance record vector, the first key feature vector, the second attendance record vector, the second key feature vector, and the verification reliability of the target identity verification mode.

[0145] In this embodiment, the server retrieves two consecutive attendance record information from the database, marks the current attendance record as the first attendance record, and marks the previous record as the second attendance record. The attendance record vector at least includes two-dimensional elements of time stamp and verification result state, and forms a standardized vector through normalization processing.

[0146] The first key feature vector and the second key feature vector can include key feature information extracted in the identity verification process. For example, for the voice recognition mode, key words can be extracted from the historical to-be-verified audio as the key feature vector; for the face recognition mode, feature points can be extracted from the face image as the key feature vector.

[0147] As an example, the verification validity of each identity verification mode in each attendance record information can be determined by the following formula 3:

[0148] Formula 3

[0149] In Formula 3, is used to represent the verification validity degree of the i-th identity authentication mode in the j-th attendance record information, is used to represent the attendance record vector corresponding to the j-th attendance record information, is used to represent the attendance record vector corresponding to the j-1-th attendance record information, is used to represent the key feature vector of the historical to-be-verified information of the i-th identity authentication mode corresponding to the j-th attendance record information, is used to represent the key feature vector of the historical to-be-verified information of the i-th identity authentication mode corresponding to the j-1-th attendance record information, is used to represent the verification reliability of the i-th identity authentication mode, norm is used to represent standardization processing, and || is used to represent the length of a vector, represents an infinitesimal greater than 0, and is used to prevent the denominator from being 0.

[0150] wherein, reflects the difference degree between adjacent attendance record information, and if the difference between adjacent attendance record vectors is large, it indicates that the attendance state may have changed greatly; represents the length difference between the two key feature vectors, which measures the change degree between the key features of adjacent historical to-be-verified information. When the difference between adjacent attendance record vectors is small, and the difference between the key feature vectors of adjacent historical to-be-verified information is large, or the identity authentication mode reliability is low, the numerator is small and the denominator is large, and the whole ratio is small, and the verification validity degree obtained after standardization is low.

[0151] Through the embodiment, the attendance record information, the historical to-be-verified information and the verification reliability of the identity authentication mode are comprehensively considered, and the verification validity degree of the identity authentication mode in the attendance record is accurately calculated. Therefore, the calculation accuracy of the verification validity degree is improved, a reliable basis is provided for subsequent construction of the verification sequence structure, and the accuracy and reliability of the entire identity authentication system are improved.

[0152] In some schemes of the present application, when the verification preference degree of each identity authentication mode in each verification sequence is determined based on the verification validity degree of each identity authentication mode in each attendance record information, it is difficult to accurately quantify, thereby leading to poor accuracy of the verification preference degree.

[0153] To this end, the present application further provides that S203 specifically can include:

[0154] The third attendance record information is screened out from each attendance record information, and the third attendance record information is used to represent the attendance record information of the target identity verification mode in the target verification sequence;

[0155] The verification preference degree of the target identity verification mode in the target verification sequence is determined by using the verification effectiveness of the target identity verification mode in each third attendance record information and the number of third attendance record information.

[0156] In this embodiment, when screening the third attendance record information, the attendance record information of the target identity verification mode in the target verification sequence is extracted by associating the verification sequence with the time dimension of the attendance record. When calculating the verification preference degree, a weighted calculation model can be constructed by combining the mean value of the verification effectiveness and the number of third attendance record information.

[0157] Specifically, first, all attendance record information of the target identity verification mode in the target verification sequence is screened out from the historical attendance record as third attendance record information. Then, the verification effectiveness of the target identity verification mode in these third attendance record information is calculated, and the number of third attendance record information is weighted as a weight factor to obtain the verification preference degree under the verification sequence.

[0158] As an example, the verification preference degree of each identity verification mode in each verification sequence can be determined by the following formula 4:

[0159] Formula 4

[0160] In formula 4, is used to represent the verification preference degree of the i-th identity verification mode in the s-th verification sequence, is used to represent the verification effectiveness of the i-th identity verification mode in the j-th attendance record information, is used to represent the number of third attendance record information of the i-th identity verification mode in the s-th verification sequence. s is used to represent the s-th verification sequence, is used to represent the maximum and minimum value normalization.

[0161] Through this embodiment, the identity verification sequence can be dynamically optimized based on historical attendance data, improving the efficiency and accuracy of the verification process. By analyzing the actual performance of each verification mode in different sequences, the system can adaptively adjust the verification strategy to better fit the actual application scenario. This data-driven method avoids the bias that may be caused by subjective judgment, making the identity verification process more scientific and reasonable. At the same time, since the optimization of the verification sequence is based on a large amount of historical data, it can adapt to the dynamic changes of the construction site environment and personnel composition, maintaining the long-term effectiveness of the verification system.

[0162] In some of the above schemes of the present application, the target user is authenticated by verifying the sequential structure, and when the voice recognition mode in the first sequence has an error, relying solely on the voice verification result to control the target user's access may have a risk of misjudgment, resulting in authorized personnel not entering the construction area and delaying the construction process.

[0163] To this end, as shown in Figure 4 S104, the present application further proposes that S104 can specifically include the following S401 to S403:

[0164] S401, comparing the second to-be-verified audio with the reference audio to obtain a voice verification result of the voice recognition mode;

[0165] S402, in the case that the voice verification result indicates that the identity authentication is passed, sending an open door instruction signal to the controller to control the opening of the gate based on the open door instruction signal, so that the target user enters or exits;

[0166] S403, in the case that the voice verification result indicates that the identity authentication is not passed, performing auxiliary identity authentication based on the verification sequence structure to control the target user's access based on the auxiliary identity authentication result.

[0167] In this embodiment, the auxiliary identity authentication process is performed based on the next sequence of the identity authentication mode in the verification sequence structure, such as fingerprint recognition or iris recognition. When the auxiliary identity authentication result is still not passed, the next sequence of the verification mode is executed in a loop until a preset stop condition is reached. The preset stop condition includes verification passing or traversing all identity authentication modes in the verification sequence structure. When the auxiliary identity authentication is passed, a manual verification process is triggered, for example, the target user's identity is confirmed by a security personnel on site, and the manual verification result is used as the basis for the final access control.

[0168] Specifically, after the target user enters the identity verification area, the first voice recognition mode in the sequence is first executed. If the voice verification result fails, the next identity verification mode in the verification sequence structure is automatically called as an auxiliary verification means. For example, if the next mode is fingerprint recognition, the fingerprint information of the target user is collected and compared with the pre-stored data. If the auxiliary verification passes, the automatic process is suspended and manual verification is triggered, for example, by transmitting real-time images to the terminal of the security personnel through the monitoring camera, and manually opening the gate after the target user's identity is confirmed by the human. If the auxiliary verification fails, the subsequent verification mode is called. When all verification modes fail, the target user is directly prohibited from entering and an alarm signal is sent. By combining automatic verification process and manual verification, the probability of false judgment can be effectively reduced, and only authorized personnel are allowed to enter high-risk areas. For example, in the case of noise interference in the construction site causing voice recognition failure, the identity is verified by fingerprint recognition, and if the fingerprint matches, the user's safety equipment compliance is confirmed by human secondary confirmation before being released.

[0169] As an example, when the target user enters the identity verification area, the server first acquires the voice audio input by the target user and compares it with the pre-stored reference audio. If the voice verification fails, the server automatically calls the next identity verification mode according to the preset verification sequence priority. For example, if the current mode is voice recognition, the next mode is iris recognition. At this time, the server activates the iris collection device, acquires the user's eye features and matches them with the iris template in the database. If the iris verification still fails, the fingerprint verification mode is called to collect the user's fingerprint information for comparison. If all verification modes fail, the server will trigger an alarm and record abnormal information. If one of the auxiliary verification modes passes, manual verification is started, and the management personnel remotely retrieve the multi-modal data of the verification process (such as voice waveform, iris image, fingerprint feature point distribution) for secondary confirmation. After the manual verification confirms the identity is legal, the server generates a temporary access credential and controls the gate to open.

[0170] Through this embodiment, the technical problem of false judgment caused by failure of a single verification mode in a multi-modal identity verification scenario is solved. By dynamically switching the verification level and combining the manual review mechanism, the identity false rejection caused by environmental interference or device failure is effectively avoided, and illegal personnel are prevented from breaking through the system defense line by forging biological characteristics. On the basis of maintaining the efficiency of automatic verification, through the cooperative mechanism of hierarchical verification and manual intervention, the reliability of personnel access control in complex construction environment is significantly improved.

[0171] In some of the above schemes of the present application, when the voice verification result indicates that the identity verification fails, the secondary identity verification is directly performed based on the verification sequence structure, but if the secondary verification result passes, it is directly released, which may have security risks of identity fraud or verification method being deceived, resulting in insufficient security of access control.

[0172] To this end, the present application further proposes that S403 can specifically include:

[0173] performing secondary identity verification based on the identity verification method at the next position in the verification sequence structure to obtain a secondary identity verification result;

[0174] In the case where the secondary identity verification result indicates that the identity verification fails, the loop is returned to perform secondary identity verification based on the identity verification method at the next position in the verification sequence structure to obtain a secondary identity verification result, until a preset stop condition is reached, the preset stop condition being that the secondary identity verification result indicates that the identity verification passes or all identity verification methods in the verification sequence structure are traversed.

[0175] In the case where the secondary identity verification result indicates that the identity verification passes, an artificial verification process is triggered, and the access of the target user is controlled based on the artificial verification result.

[0176] In the present embodiment, the preset stop condition is defined as the secondary identity verification result indicating that the identity verification passes or all identity verification methods in the verification sequence structure are traversed. When the secondary verification passes, the artificial verification process includes on-site checking of user identity information by security personnel, such as checking work certificates or inquiring about construction project details. The loop execution of secondary identity verification follows the priority order preset in the verification sequence structure, such as sequentially starting fingerprint identification or iris identification after voice recognition fails. The artificial verification result and the secondary verification result form a double confirmation mechanism, for example, when the fingerprint verification passes but the artificial verification finds that the certificate information is inconsistent, the release will be refused.

[0177] Specifically, when the voice verification fails, the server automatically switches to the next position verification method in the verification sequence structure. For example, if the second position is face recognition, the server activates the camera to collect face data and compares it with the pre-stored template. If the face recognition fails again, the third position access control card recognition is switched to. When any secondary verification method passes, the server triggers the artificial verification process, such as sending a verification request containing the user's photo and work number to the duty room. The security personnel confirm the identity authenticity through on-site observation or certificate checking, and manually input the verification result. The server only sends the opening instruction to the gate controller when the artificial verification result and the secondary verification result are double-matched. If all verification methods are not passed, the server automatically triggers an alarm and records the abnormal access event. For example, when the user fails to verify for three times in a row, the server locks the gate and uploads the abnormal log to the security management platform.

[0178] Through the embodiment, the problem of low efficiency caused by verification process redundancy in a multi-modal verification scene is solved, precise scheduling of verification resources is realized through dynamic switching of verification levels, and the risk of misjudgment caused by environmental interference or device abnormalities is effectively avoided through the linkage mechanism of manual verification and automatic verification, thereby guaranteeing the decision accuracy and execution reliability of the construction site access control system.

[0179] Based on the building construction access control method based on speech recognition provided by the application. Correspondingly, the application also provides a specific embodiment of a building construction access control system based on speech recognition.

[0180] As shown in Figure 5 , a structural diagram of a building construction access control system based on speech recognition is provided. The building construction access control system 500 based on speech recognition can include a structure acquisition module 510, an audio acquisition module 520, an audio processing module 530, and a result determination module 540.

[0181] The structure acquisition module 510 is configured to acquire a verification sequence structure when a target user enters an identity verification area, the verification sequence structure being used to represent the priority order between multiple identity verification methods.

[0182] The audio acquisition module 520 is configured to acquire a first to-be-verified audio input by the target user on site, a reference audio of the target user stored in advance, and an environmental audio of the identity verification area when the first order of the verification sequence structure is a speech recognition method.

[0183] The audio processing module 530 is configured to perform noise reduction processing on the first to-be-verified audio based on the environmental audio to obtain a second to-be-verified audio.

[0184] The result determination module 540 is configured to compare the second to-be-verified audio with the reference audio to obtain a speech verification result of the speech recognition method, so as to control the access of the target user based on the speech verification result.

[0185] In the building construction access control system based on voice recognition provided by the embodiment of the application, the verification sequence structure is arranged, when the first order is voice recognition, the first to-be-verified audio input by a target user on site, the reference audio stored in advance and the environmental audio are acquired, the first to-be-verified audio is denoised based on the environmental audio to obtain second to-be-verified audio, and then the second to-be-verified audio is compared with the reference audio to obtain a voice verification result. The voice recognition is not affected by the shielding of the protection device, the worker does not need to put down the tool in hand to perform additional operation, the situation that the card is taken to interrupt the work is avoided, and the access efficiency is improved. Meanwhile, the denoising processing can effectively reduce the environmental noise interference, the second to-be-verified audio is clearer, the voice verification result obtained by comparing the second to-be-verified audio with the reference audio is more accurate, and in combination with the verification sequence structure, various recognition modes are used scientifically and reasonably, the deficiency of the traditional method in the use of various recognition modes is made up, and the accuracy of personnel access control is comprehensively improved. Therefore, the application can effectively solve the problems of the traditional recognition mode in the building construction scene, and can improve the efficiency and accuracy of access control.

[0186] It should be noted that the present application is not limited to the specific configurations and processes described above and shown in the drawings. For the sake of brevity, detailed descriptions of known methods are omitted. In the above embodiments, several specific steps are described and shown as examples. However, the method processes of the present application are not limited to the specific steps described and shown, and those skilled in the art can make various changes, modifications and additions, or change the order of the steps, after understanding the spirit of the present application.

[0187] It should also be noted that the exemplary embodiments mentioned in the present application describe some methods or systems based on a series of steps or devices. However, the present application is not limited to the order of the above steps, that is, the steps can be performed in the order mentioned in the embodiments, or in an order different from the embodiments, or several steps can be performed simultaneously.

[0188] The above is only a specific implementation of the present application, and those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the above-described system, module and unit can refer to the corresponding process in the foregoing method embodiments, which will not be described here. It should be understood that the protection scope of the present application is not limited thereto, and any skilled person in the art can easily think of various equivalent modifications or replacements within the technical range disclosed by the present application, and these modifications or replacements should be covered within the protection scope of the present application.

Claims

1. A construction access control method based on voice recognition, characterized by, The method comprises: In the case that a target user enters an identity authentication area, an authentication sequence structure is acquired, the authentication sequence structure being used to represent a priority order between multiple identity authentication manners; In the case that a voice recognition manner is in a first order of the authentication sequence structure, a first to-be-authenticated audio input by the target user on site, a reference audio of the target user pre-stored, and an environmental audio of the identity authentication area are acquired; Based on the environmental audio, the first to-be-authenticated audio is denoised to obtain a second to-be-authenticated audio; The second to-be-authenticated audio is compared with the reference audio to obtain a voice authentication result of the voice recognition manner, so as to control the target user to enter or exit based on the voice authentication result; Before the case that a target user enters an identity authentication area, the method further comprises: Historical to-be-authenticated information of various identity authentication manners and attendance record information in a historical time period are acquired; Based on each of the historical to-be-authenticated information and corresponding attendance record information, a verification effective degree of each of the identity authentication manners in each of the attendance record information is determined; Based on the verification effective degree of each of the identity authentication manners in each of the attendance record information and a verification order of each of the identity authentication manners in each of the attendance record information, a verification preference degree of each of the identity authentication manners in each of the verification order is determined; Based on the verification preference degree of each of the identity authentication manners in each of the verification order, an authentication sequence structure is constructed; The method further comprises: Based on a difference between each of the historical to-be-authenticated information and corresponding reference authentication information, a verification possibility index of each of the historical to-be-authenticated information is determined; Based on each of the verification possibility index included in each of the identity authentication manners, a verification reliability of each of the identity authentication manners is determined; Based on each of the attendance record information and the verification reliability of each of the identity authentication manners, the verification effective degree of each of the identity authentication manners in each of the attendance record information is determined; The historical to-be-authenticated information is historical to-be-authenticated audio, and the reference authentication information is the reference audio; The method further comprises: An audio segment in the historical to-be-authenticated audio consistent with a waveform of the reference audio is determined as an effective audio segment; The effective audio segment is subjected to short-time Fourier transform processing to obtain an instantaneous phase difference and an amplitude spectrum of each time frame of the effective audio segment; Adjacent harmonic peak points in adjacent amplitude spectrums are sequentially connected to form a frequency spectrum envelope line; A maximum envelope line slope difference value is selected from envelope line slope difference values, the envelope line slope difference value being a slope difference value between adjacent frequency spectrum envelope lines; Determine a verification possibility index of the historical to-be-verified audio by using the variance of the instantaneous phase difference, the maximum envelope slope difference, and a mean value of spectral centroids, the mean value of spectral centroids being a mean value of spectral centroids corresponding to each time frame.

2. The voice recognition-based construction access control method according to claim 1, characterized by, Determine a verification reliability of each of the identity verification manners based on the verification possibility index included in each of the identity verification manners, including: Obtain an information accuracy of first historical to-be-verified information, the first historical to-be-verified information being each historical to-be-verified information corresponding to a target identity verification manner, the target identity verification manner being any one of the identity verification manners, and the information accuracy being a matching degree between the first historical to-be-verified information and the reference verification information; Perform mean value processing on the verification possibility index and the information accuracy of each of the first historical to-be-verified information respectively to obtain a mean value of verification possibility index and a mean value of information accuracy; Determine a local verification reliability of each of the first historical to-be-verified information by using the verification possibility index, the information accuracy, the mean value of verification possibility index, and the mean value of information accuracy of each of the first historical to-be-verified information; Determine a target verification reliability of the target identity verification manner based on the local verification reliability of each of the first historical to-be-verified information.

3. The voice recognition-based construction access control method according to claim 1, characterized by, Determine a verification effectiveness of each of the identity verification manners in each of the attendance record information based on each of the attendance record information and the verification reliability of each of the identity verification manners, including: Obtain a first attendance record vector of first attendance record information and a first key feature vector in second historical to-be-verified information, the second historical to-be-verified information being the historical to-be-verified information of a target identity verification manner corresponding to the first attendance record information, and the target identity verification manner being any one of the identity verification manners; Obtain a second attendance record vector of second attendance record information and a second key feature vector in third historical to-be-verified information, the second attendance record information being a previous attendance record information of the first attendance record information, and the third historical to-be-verified information being the historical to-be-verified information of a target identity verification manner corresponding to the second attendance record information; Determine a verification effectiveness of the target identity verification manner in the first attendance record information by using the first attendance record vector, the first key feature vector, the second attendance record vector, the second key feature vector, and the verification reliability of the target identity verification manner.

4. The voice recognition-based construction access control method according to claim 1, characterized by, Determine a verification preference degree of each of the identity verification manners in each of the verification order based on the verification effectiveness of each of the identity verification manners in each of the attendance record information and the verification order of each of the identity verification manners in each of the attendance record information, including: Screen third attendance record information from each of the attendance record information, the third attendance record information being used to represent the attendance record information of a target identity verification manner in a target verification order; Determine the verification preference degree of the target identity verification mode in the target verification order by the verification validity degree of the target identity verification mode in each of the third attendance record information and the quantity of the third attendance record information.

5. The voice recognition based construction access control method according to any one of claims 1 to 4, characterized in that, The comparing the second to-be-verified audio with the reference audio to obtain a voice verification result of the voice recognition mode, so as to control the access of the target user based on the voice verification result, comprises: comparing the second to-be-verified audio with the reference audio to obtain a voice verification result of the voice recognition mode; In the case that the voice verification result indicates that the identity verification is passed, a door opening instruction signal is sent to a controller, so that the controller controls the opening of a gate based on the door opening instruction signal, so that the target user can access; In the case that the voice verification result indicates that the identity verification is not passed, auxiliary identity verification is performed based on the verification sequence structure, so that the access of the target user is controlled based on the auxiliary identity verification result.

6. The voice recognition-based construction access control method according to claim 5, characterized by, The auxiliary identity verification based on the identity verification mode at the next order in the verification sequence structure to obtain the auxiliary identity verification result; In the case that the auxiliary identity verification result indicates that the identity verification is not passed, the auxiliary identity verification based on the identity verification mode at the next order in the verification sequence structure is returned to be executed to obtain the auxiliary identity verification result until a preset stop condition is reached, the preset stop condition being that the auxiliary identity verification result indicates that the identity verification is passed or all the identity verification modes in the verification sequence structure are traversed; In the case that the auxiliary identity verification result indicates that the identity verification is passed, an artificial verification process is triggered, and the access of the target user is controlled based on the artificial verification result. The system comprises:

7. A voice recognition based construction access control system, characterized by, The structure acquisition module is configured to acquire a verification sequence structure in the case that a target user enters an identity verification area, the verification sequence structure being used to represent the priority order among a plurality of identity verification modes; The audio acquisition module is configured to acquire first to-be-verified audio input by the target user on site, reference audio of the target user stored in advance, and environmental audio of the identity verification area in the case that a voice recognition mode is at a first order in the verification sequence structure; The audio processing module is configured to perform denoising processing on the first to-be-verified audio based on the environmental audio to obtain second to-be-verified audio; The result determination module is configured to compare the second to-be-verified audio with the reference audio to obtain a voice verification result of the voice recognition mode, so as to control the access of the target user based on the voice verification result; Before the structure acquisition module acquires the verification sequence structure in the case that the target user enters the identity verification area, the system further comprises: Acquire historical to-be-verified information and attendance record information of various identity verification modes in a historical time period; ​ Determine the verification effectiveness of each identity verification mode in each attendance record information based on each historical to-be-verified information and the corresponding attendance record information; Determine the verification preference of each identity verification mode in each verification order based on the verification effectiveness of each identity verification mode in each attendance record information and the verification order of each identity verification mode in each attendance record information; Construct a verification sequence structure based on the verification preference of each identity verification mode in each verification order; The method comprises the following steps: Determine the verification possibility index of each historical to-be-verified information based on the difference between each historical to-be-verified information and the corresponding reference verification information; Determine the verification reliability of each identity verification mode based on each verification possibility index included in each identity verification mode; Determine the verification effectiveness of each identity verification mode in each attendance record information based on each attendance record information and the verification reliability of each identity verification mode; The historical to-be-verified information is historical to-be-verified audio, and the reference verification information is the reference audio; The method comprises the following steps: Determine the effective audio segment in the historical to-be-verified audio as the audio segment consistent with the waveform of the reference audio; Perform short-time Fourier transform processing on the effective audio segment to obtain the instantaneous phase difference and amplitude spectrum of the effective audio segment in each time frame; Connect the harmonic peak points in adjacent amplitude spectra in sequence to form a spectral envelope line; Select the maximum envelope line slope difference from the envelope line slope differences, wherein the envelope line slope difference is the slope difference between adjacent spectral envelope lines; Determine the verification possibility index of the historical to-be-verified audio by using the variance of the instantaneous phase difference, the maximum envelope line slope difference, and the spectral centroid mean value, wherein the spectral centroid mean value is the mean value of the spectral centroid corresponding to each time frame.

Images