Medical cabinet control method and system based on voice authentication and storage medium
Through voice authentication-based technology, voice features are extracted and integrated for voiceprint matching, which solves the problem of difficult aseptic operation of medical cabinets in surgical environments, achieves contactless opening control, and improves the safety of surgical instruments.
Patent Information
- Application Number
- CN202411447817.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-16
- Publication Date
- 2025-01-21
AI Technical Summary
When using existing medical cabinets in surgical environments, due to the need to enter a password, doctors and nurses are prone to bacterial infection during sterile operations, which reduces the safety of surgical instruments.
Using voice -based control methods, the user's voice is extracted, including acoustic characteristics and emotional characteristics, and matching feature fusion and sound patterns to open medical cabinets without contact.
Effectively prevent the infection of the surgical equipment in the medical cabinet, improve the safety of the surgical equipment, and simplify the operation process in the operating room.
Smart Images
Figure CN119339723A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of smart medical technology, and in particular to a medical cabinet control method, system and storage medium based on voice authentication. Background Art
[0002] In the operating rooms of modern hospitals, with the popularization of medical cabinets (medical consumables SPD smart cabinets), intelligent management and real-time tracking of materials in the operating room have become particularly important. In order to ensure efficient workflow and material management in the operating room, how to efficiently control the opening of medical consumables SPD smart cabinets has received more and more attention.
[0003] During the use of existing medical cabinets, passwords are generally entered for control. However, due to the particularity of the surgical environment, doctors and nurses need to wear sterile surgical gowns and maintain sterile operating conditions, which can easily lead to the adsorption of bacteria and reduce the safety of surgical instruments in the medical cabinet. Summary of the Invention
[0004] The purpose of the embodiments of the present invention is to provide a medical cabinet control method, system and storage medium based on voice authentication, so as to solve the problem of low security of surgical instruments in medical cabinets in the prior art.
[0005] The embodiment of the present invention is implemented as follows: a medical cabinet control method based on voice authentication, the method comprising:
[0006] Collecting user voice to obtain voice to be authenticated, and extracting features of the voice to be authenticated to obtain acoustic features and emotional features;
[0007] Determining a preferred feature dimension according to the user's identity identifier, and performing dimension screening on the acoustic feature and the emotional feature according to the preferred feature dimension to obtain an acoustic preferred feature and an emotional preferred feature;
[0008] Fusing the acoustic optimal feature and the emotional optimal feature to obtain a user fusion feature, and performing voiceprint matching between the user fusion feature and a voiceprint database;
[0009] If the voiceprint matches, the medical cabinet is opened.
[0010] Preferably, feature extraction is performed on the speech to be authenticated to obtain acoustic features and emotional features, including:
[0011] Performing pre-emphasis processing on the speech to be authenticated according to the pre-trained acoustic model, and performing frame processing on the speech to be authenticated after the pre-emphasis processing to obtain framed speech;
[0012] Performing a fast Fourier transform on the framed speech to obtain a frequency domain representation, and calculating a power spectrum based on the frequency domain representation;
[0013] Performing linear prediction coding on the power spectrum to obtain a linear prediction signal, and calculating a difference between the power spectrum and the linear prediction signal to obtain a residual signal;
[0014] Determining a frequency spectrum according to the residual signal, and performing a cepstrum transform on the frequency spectrum to obtain the acoustic feature;
[0015] The power spectrum is filtered to obtain frequency domain energy, and the frequency domain energy is logarithmically transformed to obtain the emotional feature.
[0016] Preferably, before performing pre-emphasis processing on the speech to be authenticated according to the pre-trained acoustic model, the method further includes:
[0017] Acquiring a sample speech, and inputting the sample speech into the acoustic model for feature extraction to obtain a sample acoustic feature and a sample emotional feature;
[0018] Performing feature fusion on the sample acoustic features and the sample emotional features to obtain a sample fusion feature, and determining a model loss based on the sample acoustic features, the sample emotional features, and the sample fusion feature;
[0019] The parameters of the acoustic model are updated according to the model loss until the acoustic model converges, thereby obtaining the pre-trained acoustic model.
[0020] Preferably, after updating the parameters of the acoustic model according to the model loss until the acoustic model converges and the pre-trained acoustic model is obtained, the method further includes:
[0021] Determining the prediction accuracy of each feature dimension in the sample speech based on the sample fusion feature and the training label of the sample speech;
[0022] The preferred feature dimension is determined according to the prediction accuracy.
[0023] Preferably, the acoustic preferred feature and the emotional preferred feature are fused to obtain a user fusion feature, including:
[0024] Calculating the difference between adjacent dimensional features in the acoustic preferred feature to obtain an acoustic difference feature, and calculating the difference between adjacent dimensional features in the emotional preferred feature to obtain an emotional difference feature;
[0025] Performing a matrix multiplication operation on the acoustic difference feature and the emotional difference feature to obtain the user fusion feature.
[0026] Preferably, performing voiceprint matching on the user fusion feature and the voiceprint database includes:
[0027] Calculating the similarity between the user fusion feature and the standard feature in the voiceprint database to obtain feature similarity;
[0028] If the similarity of any of the features is greater than the similarity threshold, the voiceprint match is determined to be qualified;
[0029] If the feature similarities are all less than or equal to the similarity threshold, the voiceprint match is determined to be unqualified.
[0030] Preferably, determining the preferred feature dimension according to the user's identity identifier includes:
[0031] Determining the user's pre-collected voice according to the identity identifier, and calculating the voice similarity between the user's pre-collected voice and the sample voice;
[0032] Similar speech in the sample speech is determined according to the speech similarity, and the preferred feature dimension is determined according to the similar speech.
[0033] Another object of an embodiment of the present invention is to provide a medical cabinet control system based on voice authentication, the system comprising:
[0034] The feature extraction module is used to collect the user's voice to obtain the voice to be authenticated, and extract features from the voice to be authenticated to obtain acoustic features and emotional features;
[0035] A dimension screening module, configured to determine a preferred feature dimension according to the user's identity identifier, and perform dimension screening on the acoustic feature and the emotional feature according to the preferred feature dimension to obtain an acoustic preferred feature and an emotional preferred feature;
[0036] The voiceprint matching module is used to fuse the acoustic preferred features and the emotional preferred features to obtain user fusion features, and perform voiceprint matching on the user fusion features and the voiceprint database. If the voiceprint match is qualified, the medical cabinet is opened.
[0037] Preferably, the feature extraction module is further used to:
[0038] Performing pre-emphasis processing on the speech to be authenticated according to the pre-trained acoustic model, and performing frame processing on the speech to be authenticated after the pre-emphasis processing to obtain framed speech;
[0039] Performing a fast Fourier transform on the framed speech to obtain a frequency domain representation, and calculating a power spectrum based on the frequency domain representation;
[0040] Performing linear prediction coding on the power spectrum to obtain a linear prediction signal, and calculating a difference between the power spectrum and the linear prediction signal to obtain a residual signal;
[0041] Determining a frequency spectrum according to the residual signal, and performing a cepstrum transform on the frequency spectrum to obtain the acoustic feature;
[0042] The power spectrum is filtered to obtain frequency domain energy, and the frequency domain energy is logarithmically transformed to obtain the emotional feature.
[0043] The embodiments of the present invention can effectively determine the preferred feature dimensions through identity identification, and can effectively perform dimension screening on acoustic features and emotional features based on the preferred feature dimensions, thereby improving the accuracy of the acoustic preferred features and emotional preferred features. By fusing the acoustic preferred features and the emotional preferred features, the user fusion features corresponding to the authentication voice can be effectively obtained. By performing voiceprint matching between the user fusion features and the voiceprint database, the opening control of the medical cabinet can be performed contactlessly by adopting voiceprint matching, thereby preventing bacteria from infecting the surgical instruments in the medical cabinet and improving the safety of the surgical instruments. BRIEF DESCRIPTION OF THE DRAWINGS
[0044] Figure 1 is a flow chart of a medical cabinet control method based on voice authentication provided by the first embodiment of the present invention;
[0045] Figure 2 2 is a schematic structural diagram of a medical cabinet control system based on voice authentication provided by a second embodiment of the present invention;
[0046] Figure 3 It is a structural diagram of a terminal device provided by the third embodiment of the present invention. DETAILED DESCRIPTION
[0047] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.
[0048] In order to illustrate the technical solution of the present invention, specific embodiments are provided below.
[0049] Example 1
[0050] See also Figure 1 , is a flow chart of a medical cabinet control method based on voice authentication provided by a first embodiment of the present invention. The medical cabinet control method based on voice authentication can be applied to any device or system. The medical cabinet control method based on voice authentication includes the following steps:
[0051] Step S10: collecting the user's voice to obtain the voice to be authenticated, and extracting features from the voice to be authenticated to obtain acoustic features and emotional features;
[0052] Among them, a voice collection device is set on the medical cabinet. The voice collection device collects the user's voice in real time to obtain the voice to be authenticated. By extracting features from the voice to be authenticated, acoustic features and emotional features can be effectively obtained.
[0053] Optionally, feature extraction is performed on the speech to be authenticated to obtain acoustic features and emotional features, including:
[0054] Pre-emphasis processing is performed on the speech to be authenticated according to the pre-trained acoustic model, and the speech to be authenticated after the pre-emphasis processing is framed to obtain framed speech; wherein, by pre-emphasis processing on the speech to be authenticated, the high-frequency part of the speech to be authenticated can be effectively improved to compensate for the high-frequency information lost during the transmission process of the speech to be authenticated, thereby improving the high-frequency resolution of the speech to be authenticated. In this step, the speech to be authenticated after the pre-emphasis processing can be framed according to a preset time length to obtain framed speech;
[0055] Performing a fast Fourier transform on the framed speech to obtain a frequency domain representation, and calculating a power spectrum based on the frequency domain representation; wherein, by performing a fast Fourier transform on the framed speech, the framed speech can be effectively converted into a frequency domain signal;
[0056] Performing linear prediction coding on the power spectrum to obtain a linear prediction signal, and calculating the difference between the power spectrum and the linear prediction signal to obtain a residual signal; wherein, by performing linear prediction coding on the power spectrum, the power spectrum can be effectively linearly predicted to obtain a linear prediction signal, and by calculating the difference between the power spectrum and the linear prediction signal, the cepstrum transform of the frequency spectrum is effectively guaranteed;
[0057] Determining a frequency spectrum according to the residual signal, and performing a cepstrum transform on the frequency spectrum to obtain the acoustic feature;
[0058] The power spectrum is filtered to obtain frequency domain energy, and the frequency domain energy is logarithmically transformed to obtain the emotional feature; wherein, the frequency domain energy is logarithmically transformed to convert the frequency domain energy of the linear scale into the logarithmic scale to conform to human perception of sound, so that the feature representation is more interpretable and robust.
[0059] Furthermore, before performing pre-emphasis processing on the speech to be authenticated according to the pre-trained acoustic model, the method further includes:
[0060] Acquire a sample speech, and input the sample speech into the acoustic model for feature extraction to obtain a sample acoustic feature and a sample emotional feature; wherein the sample speech can be set according to needs, and by inputting the sample speech into the acoustic model for feature extraction, the sample acoustic feature and the sample emotional feature in the sample speech can be effectively extracted;
[0061] Performing feature fusion on the sample acoustic features and the sample emotional features to obtain a sample fusion feature, and determining a model loss based on the sample acoustic features, the sample emotional features, and the sample fusion feature; wherein a first loss is determined based on the sample acoustic features, a second loss is determined based on the sample emotional features, and a third loss is determined based on the sample fusion feature; performing a weighted operation on the first loss, the second loss, and the third loss to obtain a model loss, and the weighted values of the first loss, the second loss, and the third loss in the weighted operation can be set as required;
[0062] The parameters of the acoustic model are updated according to the model loss until the acoustic model converges, thereby obtaining the pre-trained acoustic model.
[0063] Furthermore, after updating the parameters of the acoustic model according to the model loss until the acoustic model converges and the pre-trained acoustic model is obtained, the method further includes:
[0064] Determining the prediction accuracy of each feature dimension in the sample speech based on the sample fusion feature and the training label of the sample speech; wherein, the greater the prediction accuracy, the higher the feature recognition rate of the acoustic model for the feature dimension corresponding to the prediction accuracy;
[0065] The preferred feature dimension is determined according to the prediction accuracy; wherein, among the sample acoustic features and the sample emotional features, the feature dimension corresponding to the maximum prediction accuracy is determined as the preferred feature dimension.
[0066] Step S20, determining a preferred feature dimension according to the user's identity identifier, and performing dimension screening on the acoustic feature and the emotional feature according to the preferred feature dimension to obtain an acoustic preferred feature and an emotional preferred feature;
[0067] Among them, based on the preferred feature dimension, the acoustic features and emotional features can be effectively screened, thereby improving the accuracy of the acoustic preferred features and emotional preferred features, that is, the feature information corresponding to the feature dimension with a higher recognition rate of the acoustic model can be effectively screened.
[0068] Optionally, determining a preferred feature dimension according to the user's identity identifier includes:
[0069] Determining the user's pre-collected voice according to the identity identifier, and calculating the voice similarity between the user's pre-collected voice and the sample voice; wherein the user's pre-collected voice is a standard voice pre-set by the user;
[0070] Determine the similar speech in the sample speech based on the speech similarity, and determine the preferred feature dimension based on the similar speech; wherein, in the sample speech, the speech corresponding to the maximum speech similarity is determined as the similar speech, and the feature dimension corresponding to the maximum prediction accuracy determined in the similar speech is determined as the preferred feature dimension.
[0071] Step S30, fusing the acoustic optimal feature and the emotional optimal feature to obtain a user fusion feature, and performing voiceprint matching between the user fusion feature and a voiceprint database;
[0072] Among them, by matching the user's fusion features with the voiceprint database, the opening control of the medical cabinet is carried out contactlessly by adopting the voiceprint matching method.
[0073] Optionally, the acoustic preferred feature and the emotional preferred feature are fused to obtain a user fusion feature, including: calculating the difference between adjacent dimensional features in the acoustic preferred feature to obtain an acoustic difference feature, calculating the difference between adjacent dimensional features in the emotional preferred feature to obtain an emotional difference feature, and performing matrix multiplication on the acoustic difference feature and the emotional difference feature to obtain the user fusion feature; wherein, the acoustic difference feature and the emotional difference feature are calculated in a residual manner, thereby improving the feature quality of the acoustic difference feature and the emotional difference feature.
[0074] Furthermore, performing voiceprint matching on the user fusion feature and the voiceprint database includes:
[0075] Calculating the similarity between the user fusion feature and the standard feature in the voiceprint database to obtain feature similarity; calculating the similarity between the user fusion feature and the standard feature in the voiceprint database according to the Euclidean distance formula to obtain feature similarity;
[0076] If the similarity of any of the features is greater than the similarity threshold, the voiceprint match is determined to be qualified;
[0077] If the feature similarities are all less than or equal to the similarity threshold, the voiceprint match is determined to be unqualified;
[0078] Among them, the similarity threshold can be set according to needs. If any feature similarity is greater than the similarity threshold, the user's voice authentication is determined to be qualified, that is, the user has the authority to control the medical cabinet to open. If the feature similarities are all less than or equal to the similarity threshold, the user's voice authentication is determined to be unqualified, that is, the user has no authority to control the medical cabinet to open.
[0079] Step S40: If the voiceprint matches, the medical cabinet is opened;
[0080] If the voiceprint match fails, an error message will be sent.
[0081] In this embodiment, the preferred feature dimension can be effectively determined through identity identification, and the acoustic features and emotional features can be effectively dimensionally screened based on the preferred feature dimension, thereby improving the accuracy of the acoustic preferred features and the emotional preferred features. By fusing the acoustic preferred features and the emotional preferred features, the user fusion features corresponding to the authentication voice can be effectively obtained. By matching the user fusion features with the voiceprint database, the opening control of the medical cabinet can be performed contactlessly by using voiceprint matching, thereby preventing bacteria from infecting the surgical instruments in the medical cabinet and improving the safety of the surgical instruments.
[0082] Example 2
[0083] See also Figure 2 , is a schematic structural diagram of a medical cabinet control system 100 based on voice authentication provided by a second embodiment of the present invention, comprising:
[0084] The feature extraction module 10 is used to collect the user's voice to obtain the voice to be authenticated, and perform feature extraction on the voice to be authenticated to obtain acoustic features and emotional features.
[0085] Optionally, the feature extraction module 10 is further configured to: perform pre-emphasis processing on the speech to be authenticated according to the pre-trained acoustic model, and perform frame processing on the speech to be authenticated after the pre-emphasis processing to obtain framed speech;
[0086] Performing a fast Fourier transform on the framed speech to obtain a frequency domain representation, and calculating a power spectrum based on the frequency domain representation;
[0087] Performing linear prediction coding on the power spectrum to obtain a linear prediction signal, and calculating a difference between the power spectrum and the linear prediction signal to obtain a residual signal;
[0088] Determining a frequency spectrum according to the residual signal, and performing a cepstrum transform on the frequency spectrum to obtain the acoustic feature;
[0089] The power spectrum is filtered to obtain frequency domain energy, and the frequency domain energy is logarithmically transformed to obtain the emotional feature.
[0090] Furthermore, the feature extraction module 10 is further configured to: obtain a sample speech, and input the sample speech into the acoustic model for feature extraction to obtain a sample acoustic feature and a sample emotional feature;
[0091] Performing feature fusion on the sample acoustic features and the sample emotional features to obtain a sample fusion feature, and determining a model loss based on the sample acoustic features, the sample emotional features, and the sample fusion feature;
[0092] The parameters of the acoustic model are updated according to the model loss until the acoustic model converges, thereby obtaining the pre-trained acoustic model.
[0093] Furthermore, the feature extraction module 10 is further configured to: determine the prediction accuracy of each feature dimension in the sample speech based on the sample fusion feature and the training label of the sample speech;
[0094] The preferred feature dimension is determined according to the prediction accuracy.
[0095] The dimension screening module 11 is used to determine the preferred feature dimension according to the identity identifier of the user, and perform dimension screening on the acoustic features and the emotional features according to the preferred feature dimension to obtain acoustic preferred features and emotional preferred features.
[0096] Optionally, the dimension screening module 11 is further configured to: determine the user's pre-collected voice according to the identity identifier, and calculate the voice similarity between the user's pre-collected voice and the sample voice;
[0097] Similar speech in the sample speech is determined according to the speech similarity, and the preferred feature dimension is determined according to the similar speech.
[0098] The voiceprint matching module 12 is used to fuse the acoustic preferred features and the emotional preferred features to obtain user fusion features, and perform voiceprint matching on the user fusion features with the voiceprint database. If the voiceprint match is qualified, the medical cabinet is opened.
[0099] Optionally, the voiceprint matching module 12 is further configured to: calculate the difference between adjacent dimensional features in the acoustic preferred feature to obtain an acoustic difference feature, and calculate the difference between adjacent dimensional features in the emotional preferred feature to obtain an emotional difference feature;
[0100] Performing a matrix multiplication operation on the acoustic difference feature and the emotional difference feature to obtain the user fusion feature.
[0101] Furthermore, the voiceprint matching module 12 is further configured to: calculate the similarity between the user fusion feature and the standard feature in the voiceprint database to obtain feature similarity;
[0102] If the similarity of any of the features is greater than the similarity threshold, the voiceprint match is determined to be qualified;
[0103] If the feature similarities are all less than or equal to the similarity threshold, the voiceprint match is determined to be unqualified.
[0104] In this embodiment, voiceprint features are used for identity authentication without contacting the device, which is suitable for surgical environments.
[0105] RFID, video processing, and IoT technologies are used to identify and track supplies, reducing manual inventory errors. Video surveillance information captured by cameras is analyzed to provide information on supply usage, supplementing RFID data. Combined with IoT technology, the acquired information enables interoperability between devices, improving management efficiency.
[0106] Deployment of the Voice and Voiceprint Recognition Module: The Voice and Voiceprint Recognition Module is a technology that uses human voice characteristics to authenticate identity, allowing users to complete the identity authentication process through simple voice commands. The following are the detailed implementation steps of this module:
[0107] The voiceprint registration phase is a pre-deployment phase and requires advance preparation. The preparation steps are as follows:
[0108] Step 1: Voiceprint Collection
[0109] Through a voiceprint collection interface or application, users can conveniently register their voiceprints.
[0110] The user follows the prompts to speak a specific set of phrases or sentences. These phrases should contain sufficient speech features for the voiceprint model to learn. The recording process may need to be repeated multiple times to ensure that the voiceprint model can capture the changes in speech features under different conditions.
[0111] Step 2: Preprocessing
[0112] The collected audio files are preprocessed, including noise removal, gain control, and audio format conversion, to ensure data quality. The audio files are converted into digital signals and speech features such as Mel-Frequency Cepstral Coefficients (MFCC) or other feature vectors are extracted.
[0113] Step 3: Feature extraction and training
[0114] Use machine learning or deep learning algorithms to extract features from preprocessed audio data. Train the voiceprint recognition model to enable it to distinguish between different speakers. This typically involves building a speaker model (i.e., voiceprint template) and a background model (used to capture the voice characteristics of non-target speakers).
[0115] Step 4: Create a voiceprint template
[0116] Based on the trained model, a unique voiceprint template is created for each registered user. These templates are stored in the database for use in subsequent voiceprint verification processes.
[0117] Voiceprint recognition stage
[0118] Voice acquisition equipment is installed on the SPD smart cabinet to capture clear voice commands. A voiceprint recognition algorithm is integrated, and each user pre-records their own voiceprint sample. The user authenticates through voice commands, and the system compares the voiceprint sample and grants access rights after confirming the identity. The steps in this stage are as follows:
[0119] Step 1: Real-time audio acquisition
[0120] When the user needs to verify their identity, the system activates the microphone to record the user's voice input. The user repeats the same phrase or sentence used when registering previously.
[0121] Step 2: Real-time audio preprocessing
[0122] Live audio also needs to go through a preprocessing step to ensure that its quality is suitable for voiceprint comparison.
[0123] Step 3: Real-time feature extraction
[0124] The same feature vector is extracted from the real-time audio and prepared for comparison with the existing voiceprint template.
[0125] Step 4: Voiceprint Matching
[0126] The features extracted in real time are matched with the voiceprint templates stored in the database. If the match is successful, the authentication is successful; otherwise, the authentication fails.
[0127] Step 5: Decision
[0128] The decision on whether to allow the user to access the system is made based on the voiceprint matching results. If the match is high enough, the user is granted access; otherwise, the access request is denied.
[0129] Security and privacy protection: To protect user privacy, all voiceprint data should be stored encrypted and only decrypted when necessary. The system should be able to prevent voiceprint spoofing, such as using liveness detection technology to distinguish between real human voices and recorded or synthesized voices.
[0130] User experience optimization: To improve user experience, the system should be able to respond quickly and have a certain degree of fault tolerance, such as being able to tolerate slight accent differences or environmental noise.
[0131] Through the above steps, the voiceprint recognition module can achieve efficient and secure identity authentication without interfering with the sterile environment of the operating room.
[0132] IOT module based on RFID and video processing
[0133] RFID tags and readers are integrated: Each consumable is affixed with an RFID tag that stores item information. The smart cabinet has a built-in RFID reader that automatically records when consumables enter or leave the cabinet. The reader sends this data to a central management system, providing real-time inventory status updates.
[0134] Camera video monitoring: High-definition cameras are installed inside and around smart cabinets. Video streams are transmitted to a central management system for abnormal behavior detection and inventory usage review. AI-powered video analysis aids inventory management.
[0135] IoT Platform Construction: Build a cloud-based IoT platform to enable data exchange between smart cabinets and hospital information systems. The platform receives data from RFID readers, combines it with processed video analysis data, and then feeds it back to SPD smart cabinets or hospital managers, integrating the hospital's intelligent system construction.
[0136] Voice and voiceprint recognition technology addresses the limitations of traditional identity verification methods in operating room environments, enabling contactless authentication. RFID technology enables automatic tracking of consumables, reducing human error. IoT technology connects various devices, improving the overall efficiency of material management. Camera video footage assists in monitoring, ensuring the accuracy and security of the system. In summary, this solution not only meets the requirements of aseptic operation in operating rooms, but also significantly improves the level of automation in material management, thereby effectively enhancing the work efficiency and management level of hospital operating rooms.
[0137] In this embodiment, the preferred feature dimension can be effectively determined through identity identification, and the acoustic features and emotional features can be effectively dimensionally screened based on the preferred feature dimension, thereby improving the accuracy of the acoustic preferred features and the emotional preferred features. By fusing the acoustic preferred features and the emotional preferred features, the user fusion features corresponding to the authentication voice can be effectively obtained. By matching the user fusion features with the voiceprint database, the opening control of the medical cabinet can be performed contactlessly by using voiceprint matching, thereby preventing bacteria from infecting the surgical instruments in the medical cabinet and improving the safety of the surgical instruments.
[0138] Example 3
[0139] Figure 3 This is a block diagram of a terminal device 2 provided in the third embodiment of the present application. Figure 3As shown, the terminal device 2 of this embodiment includes: a processor 20, a memory 21, and a computer program 22 stored in the memory 21 and executable on the processor 20, such as a program for a medical cabinet control method based on voice authentication. When the processor 20 executes the computer program 22, the steps of each embodiment of the medical cabinet control method based on voice authentication are implemented.
[0140] Exemplarily, the computer program 22 may be divided into one or more modules, which are stored in the memory 21 and executed by the processor 20 to implement the present application. The one or more modules may be a series of computer program instruction segments capable of performing specific functions, which are used to describe the execution process of the computer program 22 in the terminal device 2. The terminal device may include, but is not limited to, a processor 20 and a memory 21.
[0141] The processor 20 may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor.
[0142] The memory 21 may be an internal storage unit of the terminal device 2, such as a hard disk or memory of the terminal device 2. The memory 21 may also be an external storage device of the terminal device 2, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the terminal device 2. Furthermore, the memory 21 may include both an internal storage unit of the terminal device 2 and an external storage device. The memory 21 is used to store the computer program and other programs and data required by the terminal device. The memory 21 may also be used to temporarily store data that has been output or is about to be output.
[0143] In addition, the functional modules in the various embodiments of the present application may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.
[0144] If the integrated module is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Among them, the computer-readable storage medium can be non-volatile or volatile. Based on this understanding, the present application implements all or part of the process in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form, etc. The computer-readable storage medium may include: any entity or device that can carry computer program code, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium, etc. It should be noted that the content contained in computer-readable storage media can be appropriately increased or decreased according to the requirements of legislation and patent practices in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practices, computer-readable storage media do not include electrical carrier signals and telecommunications signals.
[0145] The above-described embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present application, and should all be included in the scope of protection of the present application.
Claims
1. A medical cabinet control method based on voice authentication, characterized in that: The method comprises: Collecting the user's voice to obtain the voice to be authenticated, and extracting features of the voice to be authenticated to obtain acoustic features and emotional features; Determining a preferred feature dimension according to the identity identifier of the user, and performing dimension screening on the acoustic feature and the emotional feature according to the preferred feature dimension to obtain an acoustic preferred feature and an emotional preferred feature; Fusing the acoustic optimal feature and the emotional optimal feature to obtain a user fusion feature, and performing voiceprint matching between the user fusion feature and a voiceprint database; If the voiceprint matches, the medical cabinet is opened.
2. The medical cabinet control method based on voice authentication according to claim 1, characterized in that: Extracting features of the speech to be authenticated to obtain acoustic features and emotional features, including: Pre-emphasize the speech to be authenticated according to the pre-trained acoustic model, and perform frame processing on the speech to be authenticated after the pre-emphasize processing to obtain framed speech; Performing a fast Fourier transform on the framed speech to obtain a frequency domain representation, and calculating a power spectrum based on the frequency domain representation; Performing linear prediction coding on the power spectrum to obtain a linear prediction signal, and calculating a difference between the power spectrum and the linear prediction signal to obtain a residual signal; Determine a frequency spectrum according to the residual signal, and perform a cepstrum transform according to the frequency spectrum to obtain the acoustic feature; The power spectrum is filtered to obtain frequency domain energy, and the frequency domain energy is logarithmically transformed to obtain the emotion feature.
3. The medical cabinet control method based on voice authentication as claimed in claim 2, characterized in that: Before performing pre-emphasis processing on the speech to be authenticated according to the pre-trained acoustic model, the method further includes: Acquire sample speech, and input the sample speech into the acoustic model for feature extraction to obtain sample acoustic features and sample emotional features; Performing feature fusion on the sample acoustic feature and the sample emotional feature to obtain a sample fusion feature, and determining a model loss according to the sample acoustic feature, the sample emotional feature and the sample fusion feature; The parameters of the acoustic model are updated according to the model loss until the acoustic model converges, thereby obtaining the pre-trained acoustic model.
4. The medical cabinet control method based on voice authentication as claimed in claim 3, characterized in that: The acoustic model is updated with parameters according to the model loss until the acoustic model converges to obtain the pre-trained acoustic model, and further includes: Determining the prediction accuracy of each feature dimension in the sample speech according to the sample fusion feature and the training label of the sample speech; The preferred feature dimension is determined according to the prediction accuracy.
5. The medical cabinet control method based on voice authentication according to claim 1, characterized in that: The acoustic optimal feature and the emotional optimal feature are subjected to feature fusion to obtain a user fusion feature, including: Calculating the difference between adjacent dimensional features in the acoustic optimal feature to obtain an acoustic difference feature, and calculating the difference between adjacent dimensional features in the emotional optimal feature to obtain an emotional difference feature; A matrix multiplication operation is performed on the acoustic difference feature and the emotional difference feature to obtain the user fusion feature.
6. The medical cabinet control method based on voice authentication according to claim 1, characterized in that: Performing voiceprint matching on the user fusion feature and the voiceprint database includes: Calculating the similarity between the user fusion feature and the standard feature in the voiceprint database to obtain feature similarity; If the similarity of any of the features is greater than the similarity threshold, the voiceprint match is determined to be qualified; If the feature similarities are all less than or equal to the similarity threshold, the voiceprint match is determined to be unqualified.
7. The medical cabinet control method based on voice authentication according to claim 1, characterized in that: Determining a preferred feature dimension according to the identity identifier of the user includes: Determine the user's pre-collected voice according to the identity identifier, and calculate the voice similarity between the user's pre-collected voice and the sample voice; Similar speech in the sample speech is determined according to the speech similarity, and the preferred feature dimension is determined according to the similar speech.
8. A medical cabinet control system based on voice authentication, characterized in that: The system comprises: The feature extraction module is used to collect the user's voice to obtain the voice to be authenticated, and to extract features from the voice to be authenticated to obtain acoustic features and emotional features; A dimension screening module, used to determine a preferred feature dimension according to the identity identifier of the user, and perform dimension screening on the acoustic feature and the emotional feature according to the preferred feature dimension to obtain an acoustic preferred feature and an emotional preferred feature; The voiceprint matching module is used to fuse the acoustic preferred features and the emotional preferred features to obtain user fusion features, and perform voiceprint matching on the user fusion features and the voiceprint database. If the voiceprint match is qualified, the medical cabinet is opened.
9. The medical cabinet control system based on voice authentication as claimed in claim 8, characterized in that: The feature extraction module is also used for: Pre-emphasize the speech to be authenticated according to the pre-trained acoustic model, and perform frame processing on the speech to be authenticated after the pre-emphasize processing to obtain framed speech; Performing a fast Fourier transform on the framed speech to obtain a frequency domain representation, and calculating a power spectrum based on the frequency domain representation; Performing linear prediction coding on the power spectrum to obtain a linear prediction signal, and calculating a difference between the power spectrum and the linear prediction signal to obtain a residual signal; Determine a frequency spectrum according to the residual signal, and perform a cepstrum transform according to the frequency spectrum to obtain the acoustic feature; The power spectrum is filtered to obtain frequency domain energy, and the frequency domain energy is logarithmically transformed to obtain the emotion feature.
10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.