Hearing threshold prediction system and method based on sweep frequency SFOAE fine structure

Abstract

The invention relates to a hearing threshold prediction system and method based on a sweep frequency SFOAE fine structure, and the system comprises a collection and transmission module which is used for transmitting a stimulation signal and collecting an ear canal signal. The hearing threshold value analysis and prediction mechanism comprises a hearing threshold value detection module used for detecting sweep frequency SFOAE data under each stimulation intensity based on the input stimulation intensity of a set range, constructing an SFOAE signal fine structure, extracting the SFOAE signal fine structure formed by the SFOAE signals of all the stimulation intensities under each stimulation intensity, and outputting the SFOAE signal fine structure; and calling the pre-trained network model to predict the hearing threshold. And the hearing screening module is used for detecting sweep frequency SFOAE data under the selected stimulation intensity based on the adaptively selected stimulation intensity, constructing an SFOAE signal fine structure, extracting the SFOAE signal fine structure formed by all the SFOAE signals under the selected stimulation frequency, and calling a pre-trained network model to carry out hearing screening.

Description

Technical Field

[0001] The present invention relates to the technical field of auditory system detection, and in particular to a hearing threshold prediction system and method based on a swept frequency SFOAE fine structure. Background Art

[0002] Otoacoustic emissions (OAEs) are weak audio frequencies generated in the cochlea of ​​the inner ear, transmitted through the ossicular chain and the eardrum, and released into the external auditory canal. They are part of the normal function of the human ear. Depending on the presence or absence of external stimuli, OAEs can be divided into two categories: spontaneous Otoacoustic Emissions (SOAEs) and evoked Otoacoustic Emissions (EOAEs). EOAEs are further divided into three categories: transient-evoked Otoacoustic Emissions (TEOAEs), distortion-product Otoacoustic Emissions (DPOAEs), and stimulus-frequency Otoacoustic Emissions (SFOAEs), depending on the evoked stimulus.

[0003] Since the pure-tone audiometry currently used in clinical practice is a behavioral test that requires subjective feedback from the subject, it is significantly influenced by subjective factors such as attention and cooperation. This makes this method, which requires subjective feedback, unsuitable for individuals who lack cooperation (e.g., infants and young children). Stimulus-frequency otoacoustic emissions (SFOAEs) are weak sound signals emitted by the inner ear cochlea at the same frequency as the stimulus, after the cochlea is stimulated by a single-frequency signal. SFOAEs can reflect the active mechanisms of the cochlear outer hair cells, and thus further reflect the function of the peripheral auditory system. Because the frequency of stimulus-frequency otoacoustic emissions is exactly the same as the stimulus frequency, SFOAEs have excellent frequency specificity. Furthermore, because SFOAEs can be detected in moderately and severely deaf ears at moderate and high stimulus intensities, SFOAEs have the potential to objectively and quantitatively reflect hearing thresholds, making them particularly suitable for hearing testing in individuals who lack cooperation.

[0004] The prior art discloses a portable full-function otoacoustic emission detection system, specifically a portable otoacoustic emission detection system based on a USB multimedia sound card, which realizes full-function quantitative detection and analysis of transient evoked otoacoustic emissions (TEOAEs) and distortion otoacoustic emissions (DPOAEs) signals. In the prior art, there are invention patents that disclose detection technologies and methods for estimating hearing thresholds and screening hearing conditions based on the input and output (I / O) functions of stimulation frequency otoacoustic emissions and traditional machine learning; the prior art also discloses an invention entitled "A stimulation frequency otoacoustic emission tuning curve detection and calibration system," which only discloses a detection method for the stimulation frequency otoacoustic emission suppression tuning curve and a detection technology for the calibration system, but does not involve detection technologies and methods for estimating hearing thresholds using a swept-frequency SFOAE fine structure and a machine learning network; the prior art also discloses an auditory sensitivity detection system based on stimulation frequency otoacoustic emissions, which discloses intensity sensitivity detection using the waveform shapes of each point of SFOAEs, and frequency sensitivity detection using the waveform shapes of each point of the stimulation frequency otoacoustic emission suppression tuning curve, but does not involve methods such as hearing threshold prediction using a swept-frequency SFOAE fine structure and machine learning.

[0005] In summary, some existing technologies do not perform hearing threshold detection, while others do not involve the application of swept-frequency SFOAE fine structure and machine learning to detect hearing threshold detection. Therefore, the accuracy of hearing threshold detection results needs to be further improved. Summary of the Invention

[0006] Based on this, it is necessary to provide a hearing threshold prediction system and method based on the fine structure of swept-frequency SFOAE, which can take into account both hearing threshold detection and hearing screening, and has high accuracy of hearing threshold detection results, in order to address the above technical problems.

[0007] A first aspect of the present invention provides a hearing threshold prediction system based on a swept frequency SFOAE fine structure, the system comprising:

[0008] An acquisition and transmission module, which is used to transmit stimulation signals and acquire ear canal signals;

[0009] A hearing threshold analysis and prediction mechanism, comprising a hearing threshold detection module and a hearing screening module;

[0010] The hearing threshold detection module is used to detect the swept frequency SFOAE data at each stimulation intensity based on the stimulation intensity within the set range input by the acquisition and transmission module, so as to construct the SFOAE signal fine structure at the detected stimulation intensity, and extract the SFOAE signal fine structure composed of the SFOAE signals of all stimulation intensities at each stimulation intensity, and at the same time call the pre-trained network model to predict the hearing threshold at all stimulation frequencies;

[0011] The hearing screening module is used to adaptively select the stimulation intensity based on the acquisition and transmission module, construct the SFOAE signal fine structure at the selected stimulation intensity by detecting the swept frequency SFOAE data at the selected stimulation intensity, and extract the SFOAE signal fine structure composed of the SFOAE signals at all selected stimulation frequencies, and at the same time call the pre-trained network model to screen the hearing at all selected stimulation frequencies.

[0012] In one embodiment, the acquisition and transmission module includes:

[0013] A signal sending device, the signal sending device is used to control the stimulation signal source to send a digital signal;

[0014] A signal conversion device, the signal conversion device is used to perform D / A conversion or A / D conversion on the digital signal sent or received;

[0015] a stimulation signal emitting structure, wherein the stimulation signal emitting structure is used to transmit a stimulation signal to a human ear;

[0016] A signal acquisition structure is used to acquire ear canal signals.

[0017] In one embodiment, the stimulation signal emitting structure is composed of a headphone amplifier and a micro speaker connected in sequence;

[0018] The headphone amplifier is connected to the output end of the signal conversion device. The micro-speaker includes a first electroacoustic transducer and a second electroacoustic transducer for transmitting stimulation sound and suppression sound, respectively, for inducing SFOAEs signals. The first electroacoustic transducer and the second electroacoustic transducer are both inserted into the earplug through two sound tubes, and the input ends are both connected to the headphone amplifier through a TRS interface; the micro-speaker is used to convert the analog voltage signal into an acoustic signal and transmit the acoustic signal to the ear of the subject through the earplug.

[0019] In one embodiment, the signal acquisition structure is composed of a miniature microphone and a microphone amplifier connected in sequence;

[0020] The miniature microphone includes an acoustic-electric transducer. The input end of the miniature microphone is inserted into the earplug through a transmission sound tube. The output end of the miniature microphone is connected to the input end of the microphone amplifier. The output end of the microphone amplifier is connected to the input end of the signal conversion device.

[0021] In one embodiment, the hearing threshold detection module and the hearing screening module both include:

[0022] A stimulus sound parameter setting module, wherein the stimulus sound parameter setting module is used to set stimulus sound parameters;

[0023] A sound suppression parameter setting module, wherein the sound suppression parameter setting module is used to set sound suppression parameters;

[0024] A stimulus sound signal generation module, configured to generate a corresponding digital stimulus sound signal according to set stimulus sound parameters;

[0025] A suppression sound signal generation module, the suppression sound signal generation module is used to generate a corresponding digital suppression sound signal according to the set suppression sound parameters;

[0026] A stimulation sound signal stimulation module, wherein the stimulation sound signal stimulation module is used to emit a stimulation sound signal;

[0027] The acoustic signal suppression stimulation module is configured to emit an acoustic signal suppression.

[0028] In one embodiment, the hearing threshold detection module further includes:

[0029] A hearing threshold signal detection and processing module, configured to extract stimulation frequency otoacoustic emission signals at all stimulation frequencies at different stimulation intensities from the collected ear canal signal to construct a first fine structure of the SFOAE; the curve corresponding to the first fine structure has the stimulation frequency as the abscissa and the SFOAE intensity as the ordinate;

[0030] A hearing threshold prediction module is used to predict the hearing threshold at all stimulation frequency points based on the first fine structure under different stimulation intensities through a pre-trained machine learning network model.

[0031] A second aspect of the present invention provides a hearing threshold prediction method based on the fine structure of a frequency-sweep SFOAE, which is implemented by the hearing threshold prediction system based on the fine structure of a frequency-sweep SFOAE described in the first aspect. The method comprises:

[0032] Select a detection mode; the detection mode includes a hearing threshold detection mode and a hearing screening detection mode;

[0033] Based on the selected detection mode, different stimulation signals are received through the ear canal of the subject to be tested to obtain ear canal signals, and a hearing threshold test or a hearing screening test is performed on the ear canal signals.

[0034] In one embodiment, the method of receiving different stimulation signals through the ear canal of the subject based on the selected detection mode to obtain ear canal signals, and performing hearing threshold detection or hearing screening detection on the ear canal signals includes:

[0035] When the selected mode is the hearing threshold detection mode, the stimulation sound parameters and the suppression sound parameters are set according to the specified range, and the stimulation sound signal and the suppression sound signal are transmitted to the ear of the subject to be tested to obtain the ear canal signal;

[0036] Based on the ear canal signal, a swept SFOAE fine structure at a selected stimulation intensity is constructed by detecting the full-frequency swept SFOAE data at a selected stimulation intensity; the horizontal axis of the curve corresponding to the swept SFOAE fine structure is the stimulation sound frequency, and the vertical axis is the SFOAEs intensity;

[0037] The pre-trained network model is called to predict the hearing threshold of the swept frequency SFOAE fine structure to extract the characteristic parameters of the swept frequency SFOAE fine structure; the characteristic parameters include the SFOAE amplitude, signal-to-noise ratio, phase and recall intensity at each frequency point.

[0038] In one embodiment, the method of receiving different stimulation signals through the ear canal of the subject based on the selected detection mode to obtain ear canal signals, and performing hearing threshold detection or hearing screening detection on the ear canal signals further includes:

[0039] When the selected mode is the hearing screening test mode, the stimulation sound parameters and the suppression sound parameters are set, and a plurality of specified specific stimulation intensities and a specified stimulation frequency range are input, and the stimulation sound signal and the suppression sound signal are transmitted to the ear of the subject to be tested to extract the specified SFOAE fine structure under the specified stimulation intensity; the abscissa of the curve corresponding to the specified SFOAE fine structure is the stimulation sound frequency, and the ordinate is the SFOAEs intensity;

[0040] The fine structure of the SFOAE signal composed of the SFOAE signals at all selected stimulation frequencies is extracted, and the pre-trained network model is called to screen the hearing at all selected stimulation frequencies.

[0041] In one embodiment, the pre-trained network model is composed of a support vector machine, a k-nearest neighbor algorithm, and a random forest, and is used to predict the hearing threshold of the swept-frequency SFOAE fine structure, to extract the characteristic parameters of the swept-frequency SFOAE fine structure, and to screen the hearing at all selected stimulation frequencies.

[0042] The present invention also provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, it implements the hearing threshold prediction method based on the swept frequency SFOAE fine structure described in the second aspect.

[0043] The present invention also provides a computer storage medium storing a computer program, which, when executed by a processor, implements the hearing threshold prediction based on the swept frequency SFOAE fine structure described in the second aspect.

[0044] The present invention also provides a computer program product, including a computer program, which, when executed by a processor, implements the hearing threshold prediction method based on the swept frequency SFOAE fine structure described in the second aspect.

[0045] The above-mentioned hearing threshold prediction system and method based on the fine structure of the swept frequency SFOAE has the following advantages compared with the existing technology:

[0046] (1) The present invention is based on the fine structure function of the stimulation frequency otoacoustic emission. According to the different test contents required by the test subject, different stimulation frequencies and stimulation intensities are generated based on the hearing threshold analysis and prediction system. The stimulation signal is sent through the acquisition and transmission system, and then the signal in the ear canal of the test subject is collected and input into the hearing threshold analysis and prediction system for hearing threshold detection and hearing screening, thereby realizing objective, quantitative, rapid and accurate detection of the hearing threshold of the auditory system, or realizing objective, rapid and accurate detection of the hearing screening of the auditory system.

[0047] (2) The hearing threshold detection module of the present invention is used to objectively and quantitatively extract the hearing threshold in all frequency ranges, and can objectively detect the hearing threshold in clinical practice; the hearing screening module obtains the hearing status in all frequency bands based on the adaptively selected stimulation intensity, and can realize hearing screening detection of the hearing status based on the rapid detection of the SFOAE fine structure under the selected stimulation intensity. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the technical solutions in the present invention or the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0049] Figure 1 A schematic diagram of the structure of the hearing threshold prediction system based on the swept frequency SFOAE fine structure provided by the present invention;

[0050] Figure 2 A schematic structural diagram of an embodiment of a collection and transmission system in a specific embodiment of the present invention;

[0051] Figure 3 A schematic diagram of the process of hearing threshold detection and hearing screening detection in a specific embodiment provided by the present invention;

[0052] Figure 4 This is a schematic diagram of an example of performing hearing threshold detection based on a hearing threshold test module in a specific embodiment provided by the present invention;

[0053] Figure 5 A schematic diagram of a process for predicting hearing thresholds using a machine learning network model during a hearing threshold test in a specific embodiment of the present invention;

[0054] Figure 6 A schematic diagram of the process of hearing status screening based on a machine learning network model in a specific embodiment of the present invention;

[0055] Figure 7 Schematic diagram of the first network model and the second network model in the specific embodiment provided by the present invention;

[0056] Figure 8 A schematic diagram of the flow of the hearing threshold prediction method based on the fine structure of the swept frequency SFOAE provided by the present invention;

[0057] Figure 9 This is a diagram of the internal structure of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0058] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0059] The following combination Figures 1 to 9 The present invention describes the hearing threshold prediction system and method based on the fine structure of swept frequency SFOAE.

[0060] like Figure 1 As shown, in one embodiment, a hearing threshold prediction system based on the fine structure of swept frequency SFOAE includes:

[0061] The acquisition and transmission module is used to transmit stimulation signals and acquire ear canal signals.

[0062] The hearing threshold analysis and prediction mechanism includes a hearing threshold detection module and a hearing screening module. The hearing threshold detection module is used to detect the swept-frequency SFOAE data at each stimulation intensity based on the stimulation intensity within the set range input by the acquisition and transmission module, so as to construct the fine structure of the SFOAE signal at the detected stimulation intensity, and extract the fine structure of the SFOAE signal composed of the SFOAE signals of all stimulation intensities at each stimulation intensity, and at the same time call the pre-trained network model to predict the hearing threshold at all stimulation frequencies. The hearing screening module is used to adaptively select the stimulation intensity based on the acquisition and transmission module, and construct the fine structure of the SFOAE signal at the selected stimulation intensity by detecting the swept-frequency SFOAE data at the selected stimulation intensity, and extract the fine structure of the SFOAE signal composed of the SFOAE signals at all selected stimulation frequencies, and at the same time call the pre-trained network model to screen the hearing at all selected stimulation frequencies.

[0063] In this embodiment, the acquisition and transmission module includes:

[0064] The signal sending device is used to control the stimulation signal source to send out digital signals.

[0065] Signal conversion equipment used to perform D / A conversion or A / D conversion on the digital signals sent or received.

[0066] The stimulation signal emitting structure is used to transmit stimulation signals to the human ear.

[0067] The signal acquisition structure is used to collect ear canal signals.

[0068] In this embodiment, the stimulation signal emitting structure comprises a headphone amplifier and a micro-speaker connected in sequence. The headphone amplifier is connected to the output end of the signal conversion device. The micro-speaker includes a first electroacoustic transducer and a second electroacoustic transducer for transmitting the stimulation sound and the suppression sound, respectively, for inducing the SFOAEs signal. The first electroacoustic transducer and the second electroacoustic transducer are both inserted into the earplug via two acoustic tubes, and the input ends are connected to the headphone amplifier via a TRS interface. The micro-speaker is used to convert the analog voltage signal into an acoustic signal and transmit the acoustic signal through the earplug to the ear of the subject.

[0069] In this embodiment, the signal acquisition structure consists of a miniature microphone and a microphone amplifier connected in sequence. The miniature microphone includes an acoustic-electric transducer. The input end of the miniature microphone is inserted into the earbud through a transmission sound tube. The output end of the miniature microphone is connected to the input end of the microphone amplifier, and the output end of the microphone amplifier is connected to the input end of the signal conversion device.

[0070] In this embodiment, the hearing threshold detection module and the hearing screening module both include:

[0071] The stimulus sound parameter setting module is used to set the stimulus sound parameters.

[0072] The sound suppression parameter setting module is used to set the sound suppression parameters.

[0073] The stimulation sound signal generation module is used to generate a corresponding digital stimulation sound signal according to the set stimulation sound parameters.

[0074] The suppression sound signal generating module is used to generate a corresponding digital suppression sound signal according to the set suppression sound parameters.

[0075] The stimulation sound signal stimulation module is used to send stimulation sound signals.

[0076] The suppressive sound signal stimulation module is used to send out a suppressive sound signal.

[0077] In this embodiment, the hearing threshold detection module further includes:

[0078] The hearing threshold signal detection and processing module is used to extract the stimulation frequency otoacoustic emission signals at all stimulation frequencies at different stimulation intensities from the collected ear canal signals to construct the first fine structure of the SFOAE. The curve corresponding to the first fine structure is the abscissa of the stimulation frequency and the ordinate of the SFOAE intensity.

[0079] The hearing threshold prediction module is used to predict the hearing threshold at all stimulation frequency points based on the first fine structure under different stimulation intensities through a pre-trained machine learning network model.

[0080] See also Figures 2 to 7 As shown, in a specific embodiment, the hearing threshold prediction system based on the swept frequency SFOAE fine structure provided by the present invention includes:

[0081] The acquisition and transmission system is used to transmit stimulation signals and acquire ear canal signals.

[0082] The hearing threshold analysis and prediction system is used to perform signal analysis and processing to complete hearing threshold testing or hearing screening testing.

[0083] Specifically, combined Figure 2 As shown, the acquisition and transmission system includes a signal sending device, a signal conversion device, a stimulation signal sending structure and a signal acquisition structure.

[0084] The signal sending device is used to stimulate the signal source to send a digital signal. The signal sending device can use workstation 1 to send the digital signal. The signal conversion device is used to perform A / D and D / A conversion on the signal. The signal sending device can use acquisition card 2 to achieve signal conversion. Acquisition card 2 uses an acquisition card that can be connected to workstation 1 to convert the digital signal sent by workstation 1 into an analog voltage signal. When performing detection, a portable acquisition card with a 24-bit sampling depth and a maximum sampling rate of 192kHz can be used, and connected to workstation 1 via a USB interface. Of course, the signal conversion structure can also adopt other structures and connection methods, such as acquisition card 2 connected to workstation 1 via a USB interface, which will not be repeated here.

[0085] The stimulation signal emitting structure is used to transmit stimulation signals to the human ear. The stimulation signal emitting structure may include a headphone amplifier 3 and a micro-speaker 4 connected in sequence, wherein the headphone amplifier 3 is connected to the two output ends of the acquisition card 2 to achieve power amplification and impedance matching of the two output signals of the acquisition card 2. The micro-speaker 4 includes two electro-acoustic transducers that respectively generate stimulation sound and suppression sound, which are used to induce SFOAEs signals. The two electro-acoustic transducers are inserted into the earplug through two sound tubes. The input ends of the two electro-acoustic transducers are respectively connected to the headphone amplifier 3 through interfaces. The micro-speaker 4 is used to electro-acoustically convert the analog voltage signal into an acoustic signal and transmit it to the subject's ear through the earplug. The micro-speaker 4 can adopt various products that can meet the performance indicators, such as plug-in micro-speakers.

[0086] The signal acquisition structure is used to collect otoacoustic emission signals and other signals from the external auditory canal of the human ear. The signal acquisition structure includes a miniature microphone 5 and a microphone amplifier 6 connected in sequence. To isolate the sound in the subject's external auditory canal from external sounds, the miniature speaker 4 and miniature microphone 5 can be inserted into the same soft earplug. The miniature microphone 5 includes an acoustic-to-electric transducer for collecting otoacoustic emission signals and other signals from the external auditory canal of the human ear and converting the collected acoustic signals into electrical signals. The input of the miniature microphone 5 is inserted into the earplug via an acoustic tube. The sound signal in the ear canal passes through the acoustic-to-electric transducer, where it is converted into an analog voltage signal. The output of the miniature microphone 5 is connected to the input of the microphone amplifier 6, and the output of the microphone amplifier 6 is connected to the A / D input of the acquisition card 2. The miniature microphone 5 can be a variety of products that meet the performance requirements, such as plug-in miniature microphones, and the amplification factor can be adjusted according to actual needs. Adjustable factors include but are not limited to: 0dB, 20dB, and 40dB.

[0087] Specifically, workstation 1 may also include an acquisition card driver system. This system drives the D / A port of acquisition card 2 to receive the signal from workstation 1. After receiving the signal through headphone amplifier 3 for power amplification and impedance matching, the signal is transmitted to the subject's ear through micro-speaker 4. Simultaneously, the A / D port of acquisition card 2 receives the signal sent back by microphone amplifier 6 and transmits it to the hearing threshold analysis and prediction system.

[0088] Combine Figure 3 As shown, the hearing threshold detection module system, when used for hearing threshold testing or hearing screening, first obtains the test subject's information, determines the test content, and then activates different test modules based on the test content. The hearing threshold detection module system includes a hearing threshold detection module based on the SFOAE fine structure function and a hearing screening module based on N adaptively selected intensities.

[0089] The hearing threshold detection module acquires and transmits stimulus intensities within a set range, detects the swept-frequency SFOAE data at each stimulus intensity, constructs the fine structure of the SFOAE signal at each stimulus intensity, and extracts the SFOAE fine structure composed of SFOAE signals of all stimulus frequencies at each stimulus intensity. Through a pre-trained network model, the hearing threshold at all stimulus frequencies is simultaneously predicted.

[0090] The hearing screening detection module adaptively selects the stimulation intensity through the acquisition and transmission system, constructs the fine structure of the SFOAE signal under the selected stimulation intensity by detecting the swept frequency SFOAE data under the selected stimulation intensity, and extracts the SFOAE fine structure composed of the SFOAE signals of all stimulation frequencies. Through the pre-trained network model, the hearing under all stimulation frequencies is screened.

[0091] Specifically, see Figure 4 As shown, the hearing threshold detection module includes a hearing threshold stimulation sound parameter setting module, a hearing threshold suppression sound parameter setting module, a hearing threshold stimulation sound signal generation module, a hearing threshold suppression sound signal generation module, a hearing threshold stimulation sound signal stimulation module, a hearing threshold suppression sound signal stimulation module, a hearing threshold signal detection and processing module, a hearing threshold characteristic parameter extraction module, a hearing threshold waveform display module, a hearing threshold test data display module, a hearing threshold prediction module, and a hearing threshold test result report generation and storage module.

[0092] In this embodiment, the hearing threshold stimulus sound parameter setting module is used to pre-set stimulus sound parameters, such as the frequency, intensity, and variation step size of the sweep frequency stimulus sound. The hearing threshold suppression sound parameter setting module is used to pre-set suppression sound parameters, such as the frequency and intensity of the sweep frequency suppression sound. The hearing threshold stimulus sound signal generation module is used to generate a corresponding digital stimulus sound signal based on the set stimulus sound parameters and send the corresponding signal to the hearing threshold stimulus sound signal stimulation module to send the sweep frequency stimulus sound. The hearing threshold suppression sound signal generation module is used to generate a corresponding digital suppression sound signal based on the set suppression sound parameters and send the corresponding signal to the hearing threshold suppression sound signal stimulation module to send the sweep frequency suppression sound.

[0093] The hearing threshold signal detection and processing module performs band-pass filtering, two-tone suppression filtering, online threshold rejection and offline artifact removal, superposition averaging, real-time dynamic tracking filtering on the collected ear canal signal to obtain the amplitude spectrum and phase spectrum of the swept frequency SFOAE fine structure signal. The amplitude spectrum curve of the SFOAE fine structure signal describes the relationship between the input stimulus sound frequency (horizontal axis) and the output SFOAEs intensity (vertical axis). During the specific detection, the hearing threshold stimulus sound signal stimulation module and the hearing threshold suppression sound signal stimulation module emit swept frequency stimulus sound signals and swept frequency suppression sound signals, which are D / A converted by the signal conversion structure and then sent to the subject's ear through the stimulation signal sending structure. The signal retrieval structure collects the signal collected from the subject's external auditory canal, amplifies it, and sends it to the signal conversion structure. The signal conversion structure performs A / D conversion on the signal and sends it to the hearing threshold signal detection and processing module.

[0094] In this embodiment, the hearing threshold prediction module is used to predict the hearing threshold at all stimulation frequency points based on the fine structure of the swept frequency SFOAE signal at different stimulation intensities using a pre-trained machine learning network model. Specifically,

[0095] This example uses three traditional ML algorithms—support vector machine (SVM), k-nearest neighbor (KNN), and random forest (RF)—to construct a regression model for quantitatively predicting hearing thresholds based on the fine structure of SFOAEs. A screening model for qualitatively detecting hearing loss status based on the fine structure of SFOAEs is also constructed using four ML algorithms: SVM, KNN, RF, and logistic regression (LRC). The ML process includes dataset development, feature engineering, model construction, model training, and evaluation. Selecting excellent feature engineering and appropriate prediction algorithms is key to achieving high predictive performance within traditional ML frameworks. During feature extraction, novel features of the fine structure of swept-frequency SFOAEs are extracted by considering the fine structure and the source of SFOAEs. This example utilizes traditional ML algorithms for model construction, training, and evaluation. During both model training and evaluation, the average error of 20 nested cross-validations is used as the generalization error estimate to achieve more stable model evaluation results.

[0096] The data set of the swept frequency SFOAE signal is divided as follows: first, the data is cleaned, and after training, the SFOAE fine structure that can best reflect the data characteristics is selected, including the stimulus intensity = 5-65dB SPL and 0.5-8kHz.

[0097] K-fold cross-validation (CV) is used for model training and evaluation. It divides the dataset into K folds, with each fold subset serving as the test set in turn, while the remaining K-1 folds serve as the training set. This results in K models being trained, each tested on its own test set. The errors from these K models are averaged as the CV error. Nested cross-validation (nested CV) consists of two K-fold CV loops, an outer loop and an inner loop. This effectively mitigates overfitting while fully utilizing the dataset. This example uses nested CV for model training and evaluation. In the outer loop, the dataset is divided into K folds of roughly equal sample size. One of the folds serves as the test set in turn, while the remaining K-1 folds serve as the training set. In the inner loop, the K-1 fold training set obtained in the outer loop is further divided into K-1 folds of roughly equal sample size (one fold serves as the inner loop test set, and the remaining K-1 folds serve as the inner loop training set). The inner loop's K-1-fold CV uses a grid search method to optimize hyperparameters, providing the optimal hyperparameters for the outer loop search. The final model is trained using the optimal hyperparameters on the training set provided by the outer loop, and the model is tested on the test set provided by the outer loop. This prevents information leakage and results in relatively low model performance deviation. The model error for a single nested CV is the average of the errors on the K-fold outer loop's test set. To improve the stability of the model results, this example uses the average of the errors from N nested CVs with a fixed random seed as the final estimate of the model's generalization error. A fixed random seed ensures consistent data partitioning for each nested CV, eliminating irrelevant factors for performance comparisons between models.

[0098] Regression model evaluation index: The hearing thresholds predicted by the regression model were normalized according to the following formula with an interval of 5 dB HL according to clinical rules:

[0099]

[0100] Where X is the predicted value of hearing threshold output by the regression model, Q...R are the determination coefficients, and Y ′ is the predicted value after normalization by 5dB.

[0101] In this example, the performance indicators of the regression model for quantitatively predicting hearing thresholds include mean absolute error (MAE), determination coefficient R 2 The standard error of the estimate (SE) is used as the main evaluation index and optimization index of the regression model in this example. The MAE is calculated as follows:

[0102]

[0103] Where Y i represents the actual hearing threshold of the i-th sample, Y i ′ is the predicted hearing threshold of the i-th sample, and n is the number of samples. The smaller the MAE, the more accurate the regression model prediction.

[0104] In this embodiment, the classification model evaluation indicators are suitable for evaluating the classification model for qualitatively distinguishing hearing status, including classification accuracy, F1 score, and receiver operating characteristic (ROC) curve. The classification accuracy is used as the main evaluation indicator and optimization indicator of the binary classification model, which is calculated as the proportion of the number of samples predicted correctly to the total number of samples. Precision, also known as accuracy, is the proportion of samples predicted correctly among the samples predicted to be hearing loss by the classifier; recall, also known as recall rate, is the proportion of hearing loss samples predicted correctly by the classifier to all actual hearing loss samples. The F1 score is a weighted average of the model's precision and recall, and is calculated as follows:

[0105]

[0106] In this embodiment, the ROC curve is another commonly used evaluation metric in binary classification models. It is a relationship curve plotted with the false positive rate (FPR) as the horizontal axis and the true positive rate (TPR) as the vertical axis. The FPR, also known as 1-specificity, is the proportion of samples with normal hearing that are classified as hearing loss; the TPR, also known as sensitivity, is the proportion of samples with hearing loss that are classified as hearing loss. The AUC (area under the curve) value is the area under the ROC curve. The larger the AUC, the better the classifier performance.

[0107] In this embodiment, the hearing threshold characteristic parameter extraction module is used to extract the characteristic parameters and principal components of the SFOAE fine structure function curve. The characteristic parameters are the SFOAE amplitude, signal-to-noise ratio, phase, and recall intensity parameters at each frequency point that are strongly correlated with the hearing threshold, extracted from the SFOAE fine structure function curve; the principal components are converted into an equal number of linearly uncorrelated variables through orthogonal transformation through the use of an orthogonal transformation. The principal component with the greatest correlation with the hearing threshold is then extracted according to the method used during model training and input into the hearing threshold prediction module.

[0108] The hearing threshold waveform display module dynamically displays the power spectrum waveform, baseline and noise waveforms of SFOAEs at different frequencies and stimulation intensities, as well as the SFOAE fine structure curve and noise curve at different stimulation intensities, for real-time observation of the subject's detection status and final results. Among them, the noise curve is used to observe whether the subject complies with the test requirements (the subject must be in a quiet state during the test).

[0109] The hearing threshold prediction module extracts characteristic parameters and principal components from SFOAEs data at all stimulation frequencies at different stimulation intensities and uses a pre-trained network model to predict hearing thresholds at all stimulation frequencies. The hearing threshold test result report generation and storage module displays test data at specified frequencies and different stimulation parameters, generating and saving all test results and test information for the subject.

[0110] In this embodiment, when a detailed hearing threshold test is performed on the person to be tested, the hearing threshold test module is started.

[0111] The hearing threshold detection module is executed to execute the pre-set stimulation sound parameters and suppression sound parameters in the pre-set stimulation frequency range, such as 500Hz-8kHz, and transmit the stimulation sound signal and the suppression sound signal to the acquisition and transmission system; after receiving the acquisition signal output by the acquisition and transmission system, the hearing threshold detection module detects the SFOAE data at all stimulation frequencies under each stimulation intensity to construct the SFOAE fine structure curve under the detection intensity, and then extracts the corresponding characteristic parameters and principal components through the hearing threshold characteristic parameter extraction and principal component analysis module and analyzes them. The extracted characteristic parameters include: SFOAE amplitude, signal-to-noise ratio, phase, acquisition intensity parameters, etc. at each frequency point, among which the principal component is the first two largest principal components of the frequency point that need to be predicted under the stimulation of the specified stimulation parameters, and then the hearing threshold prediction module is used to predict the hearing threshold.

[0112] Specifically, the SFOAE fine structure signal is recorded within a certain stimulation intensity and all stimulation frequency ranges, and the extracted characteristic parameters and principal components are input into a pre-trained first network model based on machine learning to determine the hearing thresholds corresponding to all stimulation frequency points and perform hearing threshold prediction; the characteristic parameters and principal components input into the first network model include but are not limited to: the spectral amplitude, noise, signal-to-noise ratio, phase, acquisition intensity parameters of the SFOAE fine structure signal at each frequency point, and the first two largest principal components of the frequency point to be predicted.

[0113] The maximum principal component of the SFOAE fine structure signal was obtained by taking the amplitude spectra and signal-to-noise ratio (S / N) spectra (amplitude spectrum minus noise) of all subjects in the frequency range of 300 to 8000 Hz as the raw data set. Separate principal component analyses were performed on the amplitude spectra and S / N spectra at specific stimulus intensities (10, 20, 30, 40, 50, 60, and 70 dB SPL), resulting in a total of seven PCAs at each frequency point. The maximum principal component analysis effectively reduces data dimensionality while minimizing information loss. In each PCA, the mean SFOAE spectrum across all subjects for that stimulus intensity and type was first subtracted from each subject's SFOAE spectrum to obtain the de-averaged SFOAE spectrum. The SFOAE spectrum data was then projected onto a new orthogonal basis (principal component space) to better account for data variance. Each subject's de-averaged SFOAE spectrum can be represented as coordinates in the orthogonally transformed principal component space, namely, the principal component score. In this example, we retain two or more principal components to explain at least 70% of the original data variance. Specifically, we set the candidate tuning parameter p to 2, 3, 4, and 5, selecting the optimal p value for each model based on its accuracy. It's worth noting that in PCA analysis, the covariance matrix is ​​trained on the training set before being applied to the test set, which is not involved in the PCA training to ensure no information leakage from the test set. This example illustrates that the principal component methods for characteristic parameters of other models are similar.

[0114] Specifically, the hearing screening module is used to perform hearing screening tests through a pre-trained machine learning-based network model, including a pre-set stimulation sound parameter setting module for screening, a suppression sound parameter setting module for screening, a stimulation sound signal generation module for screening, a suppression sound signal generation module for screening, a stimulation sound signal stimulation module for screening, a suppression sound signal stimulation module for screening, a signal detection and processing module for screening, a feature parameter extraction module for screening, a waveform display module for screening, and a test data display module for hearing screening under specific stimulation intensity.

[0115] The screening stimulus sound parameter setting module is used to set stimulus sound parameters, such as the frequency of the stimulus sound.

[0116] The screening suppression sound parameter setting module is used to set the suppression sound parameters, such as the frequency and intensity of the suppression sound.

[0117] The screening stimulation sound signal generation module and the screening suppression sound signal generation module respectively generate corresponding digital stimulation sound signals and digital suppression sound signals according to the set parameters and send the corresponding signals to the screening stimulation sound signal stimulation module and the screening suppression sound signal stimulation module.

[0118] The hearing screening detection and processing module performs band-pass filtering, two-tone suppression filtering, online threshold rejection and offline artifact removal, superposition averaging, real-time dynamic tracking filtering on the collected ear canal signal to obtain the amplitude spectrum and phase spectrum of the swept frequency SFOAE fine structure signal. The amplitude spectrum curve of the SFOAE fine structure signal describes the relationship between the input stimulus sound frequency (horizontal axis) and the output SFOAEs intensity (vertical axis). During the specific detection, the hearing screening stimulus sound signal stimulation module and the hearing screening suppression sound signal stimulation module emit swept frequency stimulus sound signals and swept frequency suppression sound signals, which are D / A converted by the signal conversion structure and then sent to the subject's ear through the stimulation signal sending structure. The signal retrieval structure collects the signal collected from the subject's external auditory canal, amplifies it, and sends it to the signal conversion structure. The signal conversion structure performs A / D conversion on the signal and sends it to the hearing screening signal detection and processing module.

[0119] The hearing screening prediction module is used to predict the hearing threshold at all stimulation frequency points based on the fine structure of the swept frequency SFOAE signal under a specified stimulation intensity through a pre-trained machine learning network model.

[0120] Specifically, a screening model for qualitatively detecting hearing loss status based on the fine structure of SFOAEs is constructed based on four ML algorithms: SVM, KNN, RF, and logistic regression (LRC). The ML process includes data sets, feature engineering, model construction, model training, and evaluation. Selecting excellent feature engineering and appropriate prediction algorithms is the core point for traditional ML frameworks to achieve high prediction performance. In the feature extraction process, the fine structure and the source of SFOAEs are considered to extract new swept-frequency SFOAEs fine structure features. This example uses traditional ML algorithms in the two parts of model construction, training, and evaluation. During the model training and evaluation process, this example uses the average error of 20 nested cross-validations as the generalization error estimate of the model to obtain more stable model evaluation results.

[0121] The feature parameter extraction module for screening is used to extract the feature parameters of the SFOAE fine structure. The feature parameters include: SFOAE amplitude, signal-to-noise ratio, phase, recall intensity, and the first two largest principal components of the frequency point that need to be predicted at each frequency point. The test data display module for screening dynamically displays the SFOAE fine structure data detected by SFOAEs under different stimulation intensities. The prediction module for screening extracts the feature parameters of the SFOAE fine structure data under a specified specific stimulation intensity, and predicts the hearing status corresponding to all stimulation frequency points through a pre-trained machine learning-based network model. The test result report generation and storage module for screening is used to generate and save all test results and test information of the subjects.

[0122] In this embodiment, when a person to be tested needs to undergo a hearing screening test, after the screening module is started, the specific calculation process is as follows:

[0123] Under the pre-set specific stimulation parameters, the stimulation sound signal and the inhibition sound signal are transmitted to the acquisition and transmission system through the N specific stimulation intensities set by the hearing screening module, and the feedback signal output by the acquisition and transmission system is input into the screening module; the screening signal detection and processing module in the screening module extracts the power spectrum signal of the stimulation frequency otoacoustic emission under the N specific stimulation intensities, such as Figure 6 As shown, the feature parameters are sent to the screening feature parameter extraction module to extract the required feature parameters, which include but are not limited to: SFOAE amplitude, signal-to-noise ratio, phase, acquisition intensity, and the first two largest principal components of the frequency point to be predicted at each frequency point; the extracted feature parameters are input into the trained second network model based on machine learning to perform hearing screening prediction.

[0124] In some embodiments, the first network model is used to predict the hearing threshold; the second network model is used to predict the hearing screening. Among them, the first network model and the second network model can adopt a network model constructed based on a machine learning algorithm or a network model constructed based on a linear regression method. The first network model and the second network model are pre-constructed and trained respectively, and pre-set in the hearing threshold analysis and prediction system or the hearing screening detection system. The network model constructed based on multivariate statistical methods includes a network model based on discriminant analysis or based on logistic regression. The network model constructed based on the machine learning algorithm includes: support vector machine, K nearest neighbor, BP neural network, random forest, decision tree and other network models. Among them, the prediction process of the hearing threshold based on the first network model and the second network model of machine learning is briefly described below:

[0125] In this embodiment, both the first and second network models utilize traditional machine learning algorithms, including support vector machines (SVM), k-nearest neighbor algorithms (KNN), decision trees (DT), random forests (RF), Adaboost, and BP neural networks (BPNN), to construct regression models to output continuous hearing thresholds. For the binary classification task of qualitatively determining hearing loss status, this example constructs a binary classification model based on SVM, KNN, DT, RF, Adaboost, and BP neural networks (BPNN) to output discrete labels of normal hearing or hearing loss.

[0126] BP neural network (Back-propagation network, BPNN) model, BP neural network model is a feedforward neural network, which uses a supervised learning technique called back propagation for training. Figure 7 As shown, the BP neural network used in this embodiment is a three-layer network consisting of an input layer, two hidden layers, and an output layer. The number of nodes in the input layer is the number of model input variables. In this embodiment, the number of nodes in the input layer of the first network model is 30, and the parameters of the input layer nodes are: SFOAE amplitude, signal-to-noise ratio, phase, recall intensity, and the first two largest principal components of the frequency point to be predicted under 5 stimulus intensities, i.e., a total of 30 nodes. The number of nodes in the input layer of the second network model in this embodiment is 12, and the parameters of the input layer nodes are: SFOAE amplitude, signal-to-noise ratio, phase, recall intensity, and the first two largest principal components of the frequency point to be predicted under 2 stimulus intensities, i.e., a total of 12 nodes. The number of nodes in the two hidden layers in this embodiment are 16 and 6 respectively. The output layer of the BP neural network model used to predict the hearing threshold has only one node, i.e., the predicted hearing threshold; while the number of nodes in the output layer of the classification model based on the BP neural network (i.e., the fourth network model in this embodiment) is 2, i.e., normal hearing or hearing loss. The training of the BP neural network model is divided into the forward propagation of the operation signal and the back propagation of the error signal. The actual output is made closer to the expected output by continuously updating the weights. The weights are fixed until the error signal is reduced to the set minimum value or reaches the set upper limit of the training steps.

[0127] The following describes the hearing threshold prediction method based on the swept-frequency SFOAE fine structure provided by the present invention. The hearing threshold prediction method based on the swept-frequency SFOAE fine structure described below and the hearing threshold prediction system based on the swept-frequency SFOAE fine structure described above can be referenced to each other.

[0128] like Figure 8 As shown, in one embodiment, a method for predicting hearing thresholds based on the fine structure of swept frequency SFOAE includes the following steps:

[0129] Step S810, selecting a detection mode; the detection mode includes a hearing threshold detection mode and a hearing screening detection mode.

[0130] Step S820: Based on the selected detection mode, different stimulation signals are received through the ear canal of the subject to be tested to obtain ear canal signals, and a hearing threshold test or a hearing screening test is performed on the ear canal signals.

[0131] In this embodiment, the hearing threshold prediction method based on the swept frequency SFOAE fine structure provided by the present invention, step S820 specifically includes the following steps:

[0132] Step S821 , when the selected mode is the hearing threshold detection mode, the stimulation sound parameters and the suppression sound parameters are set according to the specified range, and the stimulation sound signal and the suppression sound signal are transmitted to the ear of the subject to be tested to obtain the ear canal signal.

[0133] Step S822: Based on the ear canal signal, the full-band swept SFOAE data under the selected stimulation intensity are detected to construct a swept SFOAE fine structure under the selected stimulation intensity; the horizontal axis of the curve corresponding to the swept SFOAE fine structure is the stimulation sound frequency, and the vertical axis is the SFOAEs intensity.

[0134] Step S823, calling the pre-trained network model to predict the hearing threshold of the swept frequency SFOAE fine structure to extract the characteristic parameters of the swept frequency SFOAE fine structure; the characteristic parameters include the SFOAE amplitude, signal-to-noise ratio, phase and acquisition intensity at each frequency point.

[0135] In this embodiment, the hearing threshold prediction method based on the swept frequency SFOAE fine structure provided by the present invention, step S820 specifically further includes the following steps:

[0136] Step S824, when the selected mode is the hearing screening test mode, set the stimulation sound parameters and the suppression sound parameters, and input a plurality of specified specific stimulation intensities and a specified stimulation frequency range, and transmit the stimulation sound signal and the suppression sound signal to the ear of the subject to be tested to extract the specified SFOAE fine structure under the specified stimulation intensity; the horizontal coordinate of the curve corresponding to the specified SFOAE fine structure is the stimulation sound frequency, and the vertical coordinate is the SFOAEs intensity.

[0137] Step S825 , extracting the SFOAE signal fine structure composed of the SFOAE signals at all selected stimulation frequencies, and calling the pre-trained network model to screen the hearing at all selected stimulation frequencies.

[0138] It should be noted that the pre-trained network model is composed of a support vector machine, a k-nearest neighbor algorithm, and a random forest, and is used to predict the hearing threshold of the fine structure of the swept-frequency SFOAE, to extract the characteristic parameters of the fine structure of the swept-frequency SFOAE, and to screen the hearing at all selected stimulation frequencies.

[0139] Figure 9 The following is a schematic diagram of the physical structure of an electronic device. The electronic device may be a smart terminal, and its internal structure diagram may be as follows: Figure 9 As shown. The electronic device includes a processor, an internal memory, and a network interface connected via a system bus. The processor of the electronic device is used to provide computing and control capabilities. The memory of the electronic device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The network interface of the electronic device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, a hearing threshold prediction method based on the fine structure of the swept frequency SFOAE is implemented, and the method includes:

[0140] Select the test mode; the test modes include hearing threshold test mode and hearing screening test mode;

[0141] Based on the selected detection mode, different stimulation signals are received through the ear canal of the subject to be tested to obtain ear canal signals, and the ear canal signals are subjected to hearing threshold detection or hearing screening test.

[0142] Those skilled in the art will understand that Figure 9 The structure shown in the figure is merely a block diagram of a portion of the structure related to the solution of the present invention, and does not constitute a limitation on the electronic device to which the solution of the present invention is applied. The specific electronic device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0143] On the other hand, the present invention also provides a computer storage medium storing a computer program, which, when executed by a processor, implements a method for predicting hearing thresholds based on the fine structure of a swept frequency SFOAE, the method comprising:

[0144] Select the test mode; the test modes include hearing threshold test mode and hearing screening test mode;

[0145] Based on the selected detection mode, different stimulation signals are received through the ear canal of the subject to be tested to obtain ear canal signals, and the ear canal signals are subjected to hearing threshold detection or hearing screening test.

[0146] In yet another aspect, a computer program product or computer program is provided, comprising computer instructions stored in a computer-readable storage medium. A processor of an electronic device reads the computer instructions from the computer-readable storage medium, and when the processor executes the computer instructions, implements a method for predicting hearing thresholds based on a swept frequency SFOAE fine structure, the method comprising:

[0147] Select the test mode; the test modes include hearing threshold test mode and hearing screening test mode;

[0148] Based on the selected detection mode, different stimulation signals are received through the ear canal of the subject to be tested to obtain ear canal signals, and the ear canal signals are subjected to hearing threshold detection or hearing screening test.

[0149] Those skilled in the art will understand that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided by the present invention may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory.

[0150] By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), Synchronous Link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct RAM bus dynamic RAM (DRDRAM), and RAM bus dynamic RAM (RDRAM), etc.

[0151] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0152] The above-described embodiments merely illustrate several implementations of the present invention, and while their descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that a person skilled in the art would be able to make numerous variations and improvements without departing from the spirit of the present invention, all of which fall within the scope of protection of the present invention. Therefore, the scope of protection of the present invention shall be determined by the appended claims.

Claims

1. A hearing threshold prediction system based on the fine structure of swept frequency SFOAE, characterized by: The system comprises: An acquisition and transmission module, which is used to transmit stimulation signals and acquire ear canal signals; A hearing threshold analysis and prediction mechanism, comprising a hearing threshold detection module and a hearing screening module; The hearing threshold detection module is used to detect the swept frequency SFOAE data at each stimulation intensity based on the stimulation intensity within the set range input by the acquisition and transmission module, so as to construct the SFOAE signal fine structure at the detected stimulation intensity, and extract the SFOAE signal fine structure composed of the SFOAE signals of all stimulation intensities at each stimulation intensity, and at the same time call the pre-trained network model to predict the hearing threshold at all stimulation frequencies; The hearing screening module is used to adaptively select the stimulation intensity based on the acquisition and transmission module, construct the SFOAE signal fine structure at the selected stimulation intensity by detecting the swept frequency SFOAE data at the selected stimulation intensity, and extract the SFOAE signal fine structure composed of the SFOAE signals at all selected stimulation frequencies, and at the same time call the pre-trained network model to screen the hearing at all selected stimulation frequencies.

2. The hearing threshold prediction system based on the swept frequency SFOAE fine structure according to claim 1 is characterized in that: The acquisition and transmission module includes: A signal sending device, the signal sending device is used to control the stimulation signal source to send a digital signal; A signal conversion device, the signal conversion device is used to perform D / A conversion or A / D conversion on the digital signal sent or received; a stimulation signal emitting structure, wherein the stimulation signal emitting structure is used to transmit a stimulation signal to a human ear; A signal acquisition structure is used to acquire ear canal signals.

3. The hearing threshold prediction system based on the swept frequency SFOAE fine structure according to claim 2, characterized in that: The stimulation signal sending structure is composed of a headphone amplifier and a micro speaker connected in sequence; The headphone amplifier is connected to the output end of the signal conversion device. The micro-speaker includes a first electroacoustic transducer and a second electroacoustic transducer for transmitting stimulation sound and suppression sound, respectively, for inducing SFOAEs signals. The first electroacoustic transducer and the second electroacoustic transducer are both inserted into the earplug through two sound tubes, and the input ends are both connected to the headphone amplifier through a TRS interface; the micro-speaker is used to convert the analog voltage signal into an acoustic signal and transmit the acoustic signal to the ear of the subject through the earplug.

4. The hearing threshold prediction system based on the swept frequency SFOAE fine structure according to claim 3, characterized in that: The signal recovery structure is composed of a miniature microphone and a microphone amplifier connected in sequence; The miniature microphone includes an acoustic-electric transducer. The input end of the miniature microphone is inserted into the earplug through a transmission sound tube. The output end of the miniature microphone is connected to the input end of the microphone amplifier. The output end of the microphone amplifier is connected to the input end of the signal conversion device.

5. The hearing threshold prediction system based on the swept frequency SFOAE fine structure according to claim 4, characterized in that: The hearing threshold detection module and the hearing screening module both include: A stimulus sound parameter setting module, wherein the stimulus sound parameter setting module is used to set stimulus sound parameters; A sound suppression parameter setting module, wherein the sound suppression parameter setting module is used to set sound suppression parameters; A stimulus sound signal generation module, configured to generate a corresponding digital stimulus sound signal according to set stimulus sound parameters; A suppression sound signal generation module, the suppression sound signal generation module is used to generate a corresponding digital suppression sound signal according to the set suppression sound parameters; A stimulation sound signal stimulation module, wherein the stimulation sound signal stimulation module is used to emit a stimulation sound signal; The acoustic signal suppression stimulation module is configured to emit an acoustic signal suppression.

6. The hearing threshold prediction system based on the swept frequency SFOAE fine structure according to claim 5, characterized in that: The hearing threshold detection module also includes: A hearing threshold signal detection and processing module, configured to extract stimulation frequency otoacoustic emission signals at all stimulation frequencies at different stimulation intensities from the collected ear canal signal to construct a first fine structure of the SFOAE; the curve corresponding to the first fine structure has the stimulation frequency as the abscissa and the SFOAE intensity as the ordinate; A hearing threshold prediction module is used to predict the hearing threshold at all stimulation frequency points based on the first fine structure under different stimulation intensities through a pre-trained machine learning network model.

7. A hearing threshold prediction method based on the fine structure of swept frequency SFOAE, characterized in that: The method is implemented by the hearing threshold prediction system based on the swept frequency SFOAE fine structure according to any one of claims 1 to 6, and comprises: Select a detection mode; the detection mode includes a hearing threshold detection mode and a hearing screening detection mode; Based on the selected detection mode, different stimulation signals are received through the ear canal of the subject to be tested to obtain ear canal signals, and a hearing threshold test or a hearing screening test is performed on the ear canal signals.

8. The hearing threshold prediction method based on the fine structure of the swept frequency SFOAE according to claim 7, characterized in that: The method of receiving different stimulation signals through the ear canal of the subject based on the selected detection mode to obtain ear canal signals and performing hearing threshold detection or hearing screening detection on the ear canal signals includes: When the selected mode is the hearing threshold detection mode, the stimulation sound parameters and the suppression sound parameters are set according to the specified range, and the stimulation sound signal and the suppression sound signal are transmitted to the ear of the subject to be tested to obtain the ear canal signal; Based on the ear canal signal, a swept SFOAE fine structure at a selected stimulation intensity is constructed by detecting the full-frequency swept SFOAE data at a selected stimulation intensity; the horizontal axis of the curve corresponding to the swept SFOAE fine structure is the stimulation sound frequency, and the vertical axis is the SFOAEs intensity; The pre-trained network model is called to predict the hearing threshold of the swept frequency SFOAE fine structure to extract the characteristic parameters of the swept frequency SFOAE fine structure; the characteristic parameters include the SFOAE amplitude, signal-to-noise ratio, phase and recall intensity at each frequency point.

9. The hearing threshold prediction method based on the fine structure of the swept frequency SFOAE according to claim 8, characterized in that: The method of receiving different stimulation signals through the ear canal of the subject based on the selected detection mode to obtain ear canal signals, and performing hearing threshold detection or hearing screening detection on the ear canal signals, further includes: When the selected mode is the hearing screening test mode, the stimulation sound parameters and the suppression sound parameters are set, and a plurality of specified specific stimulation intensities and a specified stimulation frequency range are input, and the stimulation sound signal and the suppression sound signal are transmitted to the ear of the subject to be tested to extract the specified SFOAE fine structure under the specified stimulation intensity; the abscissa of the curve corresponding to the specified SFOAE fine structure is the stimulation sound frequency, and the ordinate is the SFOAEs intensity; The fine structure of the SFOAE signal composed of the SFOAE signals at all selected stimulation frequencies is extracted, and the pre-trained network model is called to screen the hearing at all selected stimulation frequencies.

10. The hearing threshold prediction method based on the fine structure of the swept frequency SFOAE according to claim 9, characterized in that: The pre-trained network model is composed of a support vector machine, a k-nearest neighbor algorithm, and a random forest, and is used to predict the hearing threshold of the swept-frequency SFOAE fine structure, to extract the characteristic parameters of the swept-frequency SFOAE fine structure, and to screen hearing at all selected stimulation frequencies.

Images