Double-edge cochlea dead zone rapid detection system

Abstract

The invention relates to a double-edge cochlea dead zone rapid detection system, and the system comprises an initialization test module which is used for carrying out the TEN test initialization according to a preset condition, and obtaining a series of sampling point results of a test; the model building module is used for building a response probability model of a test space based on a Gaussian process regression algorithm according to a series of tested sampling point results; the posterior probability distribution calculation module is used for establishing posterior probability distribution of the hearing threshold according to the response probability model of the test space; the sampling point determination module is used for determining an optimal next sampling point according to a series of sampling point results and posterior probability distribution of hearing thresholds and carrying out TEN testing; and the result output module is used for sending the TEN test result to the model establishment module until the test reaches a preset convergence condition, and a final cochlea dead zone detection result is obtained. The method can be widely applied to the technical field of cochlea dead zone detection in hearing loss detection.

Description

Technical Field

[0001] The present invention relates to the technical field of cochlear dead region detection in hearing loss detection, and specifically to a fast dual-edge cochlear dead region detection system based on Gaussian process TEN test active learning. Background Technique

[0002] Cochlear Dead Regions refer to regions in the cochlea where inner hair cells and / or auditory nerves cannot function properly. Therefore, information generated by the vibration of the basilar membrane in these regions cannot be transmitted to the central nervous system. In short, cochlear dead regions refer to regions in the cochlea where inner hair cells and / or auditory nerves have severe dysfunction.

[0003] For some patients, after wearing hearing aids, although the audiogram is within the assistive range of the hearing aids, there is still a situation where speech clarity is insufficient or even decreased. This is related to the dysfunction of inner hair cells and / or auditory nerves in the cochlea of the patients, indicating the existence of cochlear dead regions. No matter how strong the stimulation is received in these regions, it is difficult to convert sound signals into electrical signals. Approximately 36% to 43% of hearing aid wearers or cochlear implant recipients have cochlear dead regions of varying degrees, resulting in inaccurate hearing aid fitting results or inaccurate cochlear implantations. It can be seen that the detection of cochlear dead regions is of great significance in both hearing aid fitting and cochlear implantation.

[0004] The range of the cochlear dead region cannot be accurately identified solely from the results of pure-tone audiometry. Therefore, the current detection method for the cochlear dead region is usually the Equivalent Threshold Noise (ETN) test. That is, based on the results of pure-tone audiometry, pure-tone audiometry is performed with the superposition of a certain background noise. The threshold measured in this case is called the ETN threshold, and the range of the dead region is obtained by comparing the results of the pure-tone threshold and the ETN threshold. However, since the ETN test is a discrete-point detection and is prone to missing the dead region, it cannot well detect the edge frequency of the cochlear dead region. Therefore, in order to more accurately detect the dead-region points, a Psychophysical Tuning Curve (PTC) test can be performed. In the PTC test, the task of the subject is to judge whether a pure-tone signal exists in the presence of narrow-band masking noise. During the entire test process, the frequency (fsig) and intensity (Lsig) of the pure-tone signal remain fixed, and narrow-band masking noise with different center frequencies and intensities (fmask, Lmask) is used for each test. The detection result of the PTC is a function curve of the masking sound intensity Lmask value required for the subject to just not be able to hear the pure tone with respect to the masking sound frequency fmask. For subjects with normal hearing and subjects with hearing impairment but without a dead region, the minimum value of the PTC is near fsig, that is, at the frequency coincidence of the signal and the noise. When the signal frequency is within the dead region, the subject actually perceives the pure tone through the inner hair cells outside the dead region, or by perceiving the pure tone at the edge frequency (fe) of the dead region. Therefore, when fmask ≈ fe, the PTC will show a minimum value. Although this method can more accurately detect the edge frequency, it takes a long time (generally 30 minutes to one hour), resulting in limited application scenarios in actual production work. Summary of the Invention

[0005] Aiming at the problems that the above-mentioned existing cochlear dead region detection technologies can only achieve single-point detection and the long time-consuming leads to limitations in actual production applications, the purpose of the present invention is to provide a fast dual-edge cochlear dead region detection system based on Gaussian process ETN test active learning. This system has high detection accuracy, shorter time-consuming compared with traditional dead region detection algorithms, and can provide continuous detection results for the entire frequency band.

[0006] To achieve the above purpose, the present invention adopts the following technical solutions:

[0007] A fast dual-edge cochlear dead region detection system, comprising:

[0008] An initialization test module, configured to perform ETN test initialization according to preset conditions and obtain a series of sampling point results of the subject;

[0009] A model establishment module, configured to establish a response probability model of the test space based on the Gaussian process regression algorithm according to a series of sampling point results of the subject;

[0010] A posterior probability distribution calculation module, configured to establish a posterior probability distribution of the hearing threshold according to the response probability model of the test space;

[0011] A sampling point update module, configured to determine an optimal next sampling point according to a series of sampling point results and the posterior probability distribution of the hearing threshold, and perform a TEN test;

[0012] A result output module, configured to send the TEN test result to the model establishment module until the test reaches a preset convergence condition to obtain a final cochlear dead region detection result.

[0013] Further, the initialization test module includes:

[0014] A pure tone audiometry module, configured to obtain the pure tone audiometry result of the subject, and generate a fixed masking TEN intensity in the test space according to the audiogram of the subject;

[0015] A TEN test initialization module, configured to perform TEN test initialization based on the fixed masking TEN intensity in the test space, and obtain a series of sampling point results of the subject based on the TEN test initialization result.

[0016] Further, the model establishment module includes:

[0017] A subject response probability model establishment module, configured to establish a response probability model of the subject based on the TEN audiometry paradigm;

[0018] A frequency conversion module, configured to perform a non-linear frequency mapping conversion on the frequency of the sampling point to obtain a frequency variable;

[0019] A test space response probability model establishment module, configured to establish a response probability model of the test space based on a series of sampling point results of the subject, according to the response probability model of the subject and the frequency variable.

[0020] Further, the test space response probability model establishment module includes:

[0021] A space variable determination module, configured to establish a test space variable based on the established response probability model of the subject;

[0022] A subject response probability model establishment module, configured to establish a model of the subject response probability degree by using a Gaussian process based on the test space variable;

[0023] A posterior mean and posterior variance calculation module, configured to obtain the posterior mean and posterior variance through a Gaussian process based on a series of sampling point results of the subject;

[0024] A test space response probability model establishment module, configured to superimpose a Gaussian cumulative distribution function on the posterior mean and establish a test space response probability model of the subject.

[0025] Furthermore, the model of the degree of the test subject's response probability is expressed as:

[0026] ST(x) ~ GP(m(x), k θ (x, x))

[0027] where ST(·) is the degree value of the test subject's response probability in the test space; m(·) is the mean function, and k θ (·, ·) is the kernel function.

[0028] Furthermore, the test space response probability model establishment module includes:

[0029] The test subject response probability space establishment module is used to apply the Gaussian cumulative distribution function to the vertical axis direction of the posterior mean function to obtain the test subject response probability space;

[0030] The hearing threshold function calculation module is used to calculate the hearing threshold function based on the definition of the hearing threshold;

[0031] The model output module is used to establish the test space response probability model of the test subject based on the hearing threshold function and the test subject response probability space.

[0032] Furthermore, the posterior probability distribution of the hearing threshold is:

[0033]

[0034] where f jbark is the pure tone Bark domain frequency of the j-th sampled pulse stimulus, v is the pure tone intensity, y is the response probability of the test subject, ∑ * is the posterior variance of the Gaussian process, v * is the hearing threshold, σ p is the hyperparameter, and is the hearing threshold function.

[0035] Furthermore, the sampling point update module includes:

[0036] The objective function determination module is used to determine the objective function based on a series of sampling point results and posterior probabilities of the test subject;

[0037] The next optimal sampling point determination module is used to transform the objective function based on the symmetry of mutual information and take the maximum value of the product of the result and the posterior variance of the Gaussian process as the next optimal sampling point.

[0038] Furthermore, the next optimal sampling point is expressed as:

[0039]

[0040] where (f *bark , v * ) is the next stimulus that can provide the maximum data information for ; is the hearing threshold function; v is the pure tone intensity, y is the response probability of the subject, ∑ * is the posterior variance of the Gaussian process, D n is the existing sampling point, and f bark is the frequency variable.

[0041] Further, in the result output module, the preset convergence condition is to satisfy any of the following conditions:

[0042] The first one: The difference between the current test result and the previous test result is within 2 dB for 3 consecutive times or more.

[0043] The second one: The number of sampling points reaches 64.

[0044] Due to the above technical solutions adopted by the present invention, it has the following advantages:

[0045] 1. The present invention first proposes to apply the Gaussian process model in the detection of cochlear dead zones, and superimposes prior data, reducing the detection space. Compared with the traditional TEN test and the swept-frequency TEN test, the speed of cochlear dead zone detection is improved.

[0046] 2. Compared with the existing Fast PTC and active learning PTC technologies, the cochlear dead zone technology of the present invention can achieve continuous cochlear dead zone detection across the entire frequency band, while the existing technologies can only record single-point data at the edge of the dead zone.

[0047] 3. Since the test paradigm is a single-frequency point test rather than a swept-frequency paradigm, it can effectively avoid the situations of "missing tracking" and "overtracking" of the stimulus sound by the subject, and the test results have better accuracy and stability.

[0048] The present invention can be widely applied to the technical field of cochlear dead zone detection in hearing loss detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] By reading the following detailed description of the preferred embodiments, various other advantages and benefits will become clear to those of ordinary skill in the art. The drawings are only for the purpose of showing the preferred embodiments and are not considered to be a limitation of the present invention. Throughout the drawings, the same reference numerals are used to represent the same components. In the drawings:

[0050] Figure 1 is a schematic diagram of the dual-edge cochlear dead zone rapid detection system provided by the embodiment of the present invention;

[0051] Figure 2The conversion from linear frequency to Bark domain frequency provided by the embodiments of the present invention;

[0052] Figure 3 The posterior mean matrix obtained by the Gaussian process provided by the embodiments of the present invention;

[0053] Figure 4 The cumulative Gaussian distribution function model provided by the embodiments of the present invention;

[0054] Figure 5 The hearing response probability rectangle with the Gaussian cumulative distribution function superimposed on the posterior mean provided by the embodiments of the present invention;

[0055] Figure 6 The mutual information result calculated based on the posterior mean and posterior variance of the Gaussian process provided by the embodiments of the present invention;

[0056] Figure 7 The posterior variance matrix obtained by the Gaussian process provided by the embodiments of the present invention;

[0057] Figure 8 The final optimized objective function matrix obtained based on the posterior variance matrix and the mutual information matrix provided by the embodiments of the present invention;

[0058] Figure 9 The final detection result of the double-edge dead zone detection algorithm based on Gaussian process TEN test active learning provided by the embodiments of the present invention;

[0059] Figure 10 The result convergence graph provided by the embodiments of the present invention. Detailed implementation manners

[0060] To make the objectives, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are some but not all of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the described embodiments of the present invention fall within the scope of protection of the present invention.

[0061] It should be noted that the terms used herein are only for describing specific implementation manners and are not intended to limit the exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular forms are also intended to include the plural forms. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, they specify the presence of the described features, steps, operations, devices, components and / or their combinations.

[0062] In some embodiments of the present invention, a dual-edge cochlear dead zone rapid detection system is provided, which includes: initializing according to preset conditions and obtaining a series of sampling point results; based on the series of sampling point results, establishing a response probability model of the test space using the Gaussian process regression algorithm; according to the response probability model of the test space, establishing a posterior probability distribution of the hearing threshold; according to the series of sampling point results and the posterior probability distribution of the hearing threshold, determining the optimal next sampling point and performing cochlear dead zone detection; repeating the above steps until the preset convergence condition is reached to obtain the final cochlear dead zone detection result. The detection of the present invention has high accuracy, takes less time compared to traditional dead zone detection algorithms, and can provide continuous detection results for the entire frequency band.

[0063] Embodiment 1

[0064] As Figure 1 shown, the present invention provides a dual-edge cochlear dead zone rapid detection system, which includes the following steps:

[0065] An initialization test module, configured to initialize the TEN test according to preset conditions and obtain a series of sampling point results of the subject;

[0066] A model establishment module, configured to establish a response probability model of the test space based on a series of sampling point results of the subject using the Gaussian process regression algorithm;

[0067] A posterior probability distribution calculation module, configured to establish a posterior probability distribution of the hearing threshold according to the response probability model of the test space;

[0068] A sampling point determination module, configured to determine the optimal next sampling point and perform the TEN test according to a series of sampling point results and the posterior probability distribution of the hearing threshold;

[0069] A result output module, configured to send the TEN test result to the model establishment module until the test reaches the preset convergence condition to obtain the final cochlear dead zone detection result.

[0070] Furthermore, the initialization test module includes:

[0071] A pure tone audiometry module, configured to obtain the pure tone audiometry result of the subject and generate a fixed masking TEN intensity in the test space according to the audiogram of the subject, where the TEN intensity is the hearing threshold corresponding to the pure tone audiometry frequency point;

[0072] A TEN test initialization module, configured to initialize the TEN test based on the fixed masking TEN intensity in the test space and obtain a series of sampling point results of the subject based on the TEN test initialization result.

[0073] Further, in the TEN test initialization module, performing TEN test initialization means giving a test stimulus signal, where the stimulus signal is the simultaneous superposition of TEN noise and pulsed pure tone in the ipsilateral ear, and the subject responds to the given stimulus through a responder according to whether they can hear the pulsed pure tone (pressing indicates responding that they heard, not pressing indicates responding that they did not hear). Specifically, when initializing the stimulus signal, starting from 1000 Hz, sampling is performed upward at a preset frequency interval. For example, the sampled frequency points are successively 1500 Hz, 2000 Hz, 3000 Hz, 4000 Hz; then starting from 1000 Hz, sampling is performed downward at a preset frequency interval. For example, the sampled frequency points are successively 750 Hz, 500 Hz; the initial sampling intensity at each frequency point is randomly increased by 10 - 15 dB based on the pure tone audiometry result (TEN intensity) at the corresponding frequency point, and there is an increase prior. When the pure tone audiometry result (TEN intensity) at the corresponding frequency point drops by 10 dB, the sampling result of the subject is "no response".

[0074] Assume that after TEN test initialization, a series of sampling point results are obtained, which are represented as follows:

[0075]

[0076] In the formula, D n is the existing sampling point set, d j is the result of each single sampling point, f jbark is the Bark domain frequency of the pulsed stimulus pure tone at the j - th sampling, v j is the intensity of the pulsed stimulus pure tone, y j is the binary result of whether a response was made, "2" indicates that the responder was pressed to make a response, "-2" indicates that the responder was not pressed, and x j is the test space variable, which is the abbreviation collection of the stimulus signal information.

[0077] Further, the model establishment module includes:

[0078] The subject response probability model establishment module is used to establish the subject's response probability model based on the TEN audiometry paradigm;

[0079] The frequency conversion module is used to perform non - linear frequency mapping conversion on the frequencies of the sampling points to obtain frequency variables;

[0080] The test space response probability model establishment module is used to establish the response probability model of the test space based on a series of sampling point results of the subject, according to the subject's response probability model and the frequency variables.

[0081] Further, in the subject response probability model establishment module, establishing the subject's response probability model based on the TEN audiometry paradigm is expressed as:

[0082]

[0083] In the formula, f is the frequency, with the unit of Hz; v is the pure tone intensity, with the unit of dBHL.

[0084] Furthermore, in the frequency conversion module, since the human ear's perception of different frequencies is not linear but logarithmic, based on different definitions, there are currently various non-linear frequency mapping methods in practice. In this embodiment, the frequency of the sampling points is converted by the mapping in the Bark domain to obtain a frequency variable, that is:

[0085]

[0086] In the formula, f is the frequency, with the unit of Hz.

[0087] Assume that the functional expression of the hearing threshold is For the variable f is mapped, and there is:

[0088]

[0089] Furthermore, the test space response probability model establishment module includes:

[0090] The space variable determination module is used to establish a test space variable x = (f bark , v) based on the established response probability model of the subject;

[0091] The subject response probability model establishment module is used to establish a model of the subject response probability degree based on the test space variable x by using the Gaussian process;

[0092] The posterior mean and posterior variance calculation module is used to obtain the posterior mean and posterior variance through the Gaussian process based on a series of sampling point results of the subject;

[0093] The test space response probability model establishment module is used to superimpose the Gaussian cumulative distribution function on the posterior mean and establish a test space response probability model of the subject.

[0094] Furthermore, in the subject response probability model establishment module, the model of the subject response probability degree can be expressed as:

[0095] ST(x) ~ GP(m(x), k θ (x, x)) (5)

[0096] In the formula, ST(·) is the degree value of the subject response probability in the test space, and the larger this value is, the greater the probability of response; m(·) is the mean function, and k θ (·, ·) (covariance function) is the kernel function.

[0097] As can be seen from Equation (5), the Gaussian process is completely determined by the mean function and the kernel function. In this embodiment, the mean function is set to 0, and the kernel function is composed of a linear kernel and a squared exponential kernel, that is:

[0098] k θ (x,x) = k L (v,v) + k SE (f bark ,f bark ) (6)

[0099] It is intuitive to know that the greater the loudness, the greater the probability of the subject's response. Therefore, in order to represent the characteristic of increasing values along the longitudinal (loudness axis), a linear kernel is used to act on the loudness variable v:

[0100] k L (v,v) = Lv T v (7)

[0101] In the formula, L is a linear change constant, and in this embodiment, L is set to 0.8.

[0102] In order to represent the non-linear smooth change characteristic along the transverse (frequency axis), a squared exponential kernel function is used to act on the frequency variable f bark :

[0103]

[0104] Furthermore, in the posterior mean and posterior variance calculation module, the calculation formulas for the posterior mean and posterior variance are:

[0105]

[0106] In the formula, x is each discrete data point of the test, x * is the entire continuous space of observation, μ * is the posterior mean of the Gaussian process, ∑ * is the posterior variance of the Gaussian process, I is the identity matrix, σ 2 is the hyperparameter representing the variance or scale factor of the signal by the squared exponential kernel, which controls the amplitude of the kernel function and affects the overall magnitude of the function value.

[0107] Furthermore, the test space response probability model establishment module includes:

[0108] The subject response probability space establishment module is used to apply the Gaussian cumulative distribution function in the longitudinal axis (loudness axis) direction of the posterior mean function to obtain the subject response probability space, so as to convert the probability degree of the subject's response into the subject response probability;

[0109] The hearing threshold function calculation module is used to calculate the hearing threshold function based on the definition of the hearing threshold.

[0110] A model output module, configured to establish a test space response probability model of a subject based on a hearing threshold function and a subject response probability space.

[0111] Further, the subject response probability space is expressed as:

[0112]

[0113] Further, in this embodiment, the probability point (0.5) at which the subject can just hear is used as the hearing threshold v * :

[0114] v * = v| P(y|x*)＝0.5 (11)

[0115] In the formula, σ p represents an uncertainty hyperparameter of the subject response itself. Since the sampled response values range from -2 to +2, the value 0 is considered the mean value of the response. Based on this, the hearing threshold function can be obtained:

[0116]

[0117] Further, the test space response probability model of the subject is expressed as:

[0118]

[0119] Further, in the posterior probability distribution calculation module, based on the response probability model of the test space, the posterior probability distribution of the hearing threshold is obtained as:

[0120]

[0121] Further, the sampling point update module includes:

[0122] An objective function determination module, configured to determine an objective function based on a series of sampling point results and posterior probabilities of the subject.

[0123] A next optimal sampling point determination module, configured to transform the objective function based on the symmetry of mutual information and use the maximum value of the product of the transformed objective function and the posterior variance of the Gaussian process as the next optimal sampling point.

[0124] Further, for the objective function determined by the objective function determination module, given the existing sampling points D n and the posterior probability Our goal is to find a next stimulus (f , v *bark , v * ) that can provide the maximum data information for Stimulation test of "expected posterior entropy decrease":

[0125]

[0126] In the formula, H(·|·) represents conditional Shannon entropy.

[0127] 4.2) Transform the objective function based on the symmetry of mutual information, and take the maximum value of the result of multiplying it by the posterior variance of the Gaussian process as the next optimal sampling point.

[0128] It can be seen from Equation (15) that the objective function is equivalent to and the conditional mutual information of y: Due to the symmetry of mutual information, "maximizing the expected posterior entropy decrease" in the formula and y can be interchanged to obtain an expression that is easier to calculate:

[0129]

[0130] where the Shannon conditional entropy can be approximated as:

[0131]

[0132] In the formula, σ n represents a hyperparameter of the variance uncertainty of the hearing threshold probability at a certain frequency point. The larger it is, the larger the range of the hearing threshold uncertainty corresponding to the frequency point.

[0133] The second term in the conditional mutual information can be well approximated by replacing the binary entropy with the squared exponential function proposed by Houlsby:

[0134]

[0135] In the formula, C is an intermediate variable.

[0136] Therefore, it can be obtained that:

[0137]

[0138] In order to further expand the relative value of the test space uncertainty, in this embodiment, the maximum value of the result of multiplying the conditional mutual information by the posterior variance of the Gaussian process is used as the next optimal sampling point:

[0139]

[0140] Furthermore, in the result output module, the preset convergence condition is to satisfy any of the following conditions:

[0141] The first type: The difference between the current test result and the previous test result is within 2 dB for 3 consecutive times or more.

[0142] The second type: The number of sampling points reaches 64.

[0143] Example 2

[0144] This example further introduces the active learning dual-edge cochlear dead zone detection system based on Gaussian process TEN testing provided in Example 1. The specific detection process is as follows:

[0145] 1) Initialization: Based on preset conditions, perform cochlear dead zone detection on the subject to obtain a series of sampling points.

[0146] 2) After initialization, based on the results of a series of sampling points, establish a response probability model for the test space using the Gaussian process regression algorithm.

[0147] As Figure 2 shown, map the frequency of the sampling points through the linear frequency to the frequency in the Bark domain.

[0148] As Figure 3 shown, it is the posterior mean result graph of Gaussian process regression. The "×" in the graph represents that the subject responded to this test point, and the "○" represents that the subject did not respond to this test point.

[0149] As Figure 4 shown, superimpose the Gaussian cumulative distribution function on the posterior mean to obtain the subject response probability space.

[0150] As Figure 5 shown, it is the result of the response probability space.

[0151] 3) Based on the response probability model of the test space, obtain the posterior probability distribution of the hearing threshold.

[0152] 4) Determine the optimal next sampling point.

[0153] As Figure 6 shown, it is the result of the mutual information.

[0154] As Figure 7 shown, it is the result of the posterior variance.

[0155] As Figure 8 shown, it is the result of the objective function of the product of the final posterior variance and mutual information.

[0156] 5) Repeat steps 2) to 4) until the set convergence condition is reached.

[0157] As Figure 9As shown, it is the result of the double-edge cochlear dead zone detection algorithm for the final detection. The blue continuous solid line is the result of the continuous frequency band detected by this algorithm, and the red dashed line is the TEN intensity of the corresponding frequency points.

[0158] As Figure 10 shown, it is the result convergence process diagram of the algorithm proposed by the present invention. The value of each point represents the difference between the current detection result and the previous detection result. The smaller the value, the closer it is to the convergence boundary. The gray area below the red dashed line is the convergence space.

[0159] After using the technical solution of the present invention, for the same test target (detection of cochlear dead zone), the traditional TEN test and PTC test altogether take nearly 20 minutes, and the test result can only be a single-point result. The present invention can ensure that the test result converges within 4 minutes, and the test result is a continuous result within the entire frequency band. The present invention not only ensures the accuracy of the test result but also shortens the test time.

[0160] Those skilled in the art should understand that the embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0161] The present invention is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to embodiments of the present invention. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of flows and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to the processors of general-purpose computers, special-purpose computers, embedded processors, or other programmable data processing devices to generate a machine, such that the instructions executed by the processors of the computer or other programmable data processing devices generate means for implementing the functions specified in Figure 1 one or more flows and / or blocks Figure 1 one or more blocks.

[0162] These computer program instructions can also be stored in a computer-readable memory capable of guiding a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory generate a manufactured article including instruction means, and the instruction means implements the functions specified in Figure 1 one or more flows and / or blocks Figure 1 one or more blocks.

[0163] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus, so that a series of operation steps are performed on the computer or other programmable apparatus to generate a computer-implemented process, thereby providing instructions for implementing the steps specified in one process or a plurality of processes and / or blocks Figure 1 one process or a plurality of processes and / or blocks Figure 1 or functions specified in a plurality of blocks.

[0164] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the above embodiments, those of ordinary skill in the art should understand that: modifications or equivalent replacements can still be made to the specific embodiments of the present invention. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention shall be covered by the protection scope of the claims of the present invention.

Claims

1. A dual-edge cochlear dead region rapid detection system, characterized in that Including: An initialization test module, configured to perform TEN test initialization according to preset conditions and obtain a series of sampling point results of the subject; A model establishment module, configured to establish a response probability model of the test space based on a series of sampling point results of the subject and the Gaussian process regression algorithm; A posterior probability distribution calculation module, configured to establish a posterior probability distribution of the hearing threshold according to the response probability model of the test space; A sampling point update module, configured to determine the optimal next sampling point according to a series of sampling point results and the posterior probability distribution of the hearing threshold and perform a TEN test; A result output module, configured to send the TEN test results to the model establishment module until the test reaches the preset convergence condition to obtain the final cochlear dead zone detection result.

2. A double-edge cochlear dead zone rapid detection system according to claim 1, characterized in that, The initialization test module includes: A pure tone audiometry module, configured to obtain the pure tone audiometry results of the subject and generate a fixed masking TEN intensity in the test space according to the audiogram of the subject; A TEN test initialization module, configured to perform TEN test initialization based on the fixed masking TEN intensity in the test space and obtain a series of sampling point results of the subject based on the TEN test initialization results.

3. A dual-edge cochlear dead zone rapid detection system according to claim 1, characterized in that, The model establishment module includes: A subject response probability model establishment module, configured to establish a response probability model of the subject based on the TEN audiometry paradigm; A frequency conversion module, configured to perform a non-linear frequency mapping conversion on the frequency of the sampling points to obtain a frequency variable; A test space response probability model establishment module, configured to establish a response probability model of the test space based on a series of sampling point results of the subject, the response probability model of the subject, and the frequency variable.

4. A double-edge cochlear dead zone rapid detection system according to claim 3, characterized in that, The test space response probability model establishment module includes: A space variable determination module, configured to establish a test space variable based on the established response probability model of the subject; A subject response probability model establishment module, configured to establish a model of the degree of the subject's response probability using the Gaussian process based on the test space variable; A posterior mean and posterior variance calculation module, configured to obtain the posterior mean and posterior variance through the Gaussian process based on a series of sampling point results of the subject; A test space response probability model establishment module, configured to superimpose the Gaussian cumulative distribution function on the posterior mean and establish a response probability model of the test space of the subject.

5. A double-edge cochlear dead zone rapid detection system according to claim 4, characterized in that, The model of the degree of the subject's response probability is expressed as: ST(x) ~ GP(m(x), k θ (x, x)) where ST(·) is the degree value of the response probability of the subject in the test space; m(·) is the mean function, and k θ (·,·) is the kernel function.

6. The rapid detection system for double-edge cochlear dead zones according to claim 4, wherein, The test space response probability model establishment module includes: A subject response probability space establishment module, configured to apply the Gaussian cumulative distribution function to the vertical axis direction of the posterior mean function to obtain a subject response probability space; A hearing threshold function calculation module, configured to calculate a hearing threshold function based on the definition of the hearing threshold; A model output module, configured to establish a response probability model of the test space of the subject based on the hearing threshold function and the subject response probability space.

7. A dual-edge cochlear dead zone rapid detection system according to claim 1, wherein The posterior probability distribution of the hearing threshold is: where f jbark is the Bark domain frequency of the pulsed stimulus pure tone for the j-th sampling, v is the pure tone intensity, y is the response probability of the subject, and ∑ * is the posterior variance of the Gaussian process, v * is the hearing threshold, σ p is the hyperparameter, and is the hearing threshold function.

8. A double-edge cochlear dead zone rapid detection system according to claim 1, characterized in that The sampling point update module includes: An objective function determination module, configured to determine an objective function based on a series of sampling point results of the subject and the posterior probability; A next optimal sampling point determination module, configured to convert the objective function based on the symmetry of mutual information and use the maximum value of the product of the converted objective function and the posterior variance of the Gaussian process as the next optimal sampling point.

9. The rapid detection system for double-edge cochlear dead zones according to claim 8, characterized in that, The next optimal sampling point is expressed as: where (f *bark , v * ) is the next stimulus that can provide the maximum data information for ; is the hearing threshold function; v is the pure tone intensity, y is the response probability of the subject, ∑ * is the posterior variance of the Gaussian process, D n is the existing sampling point, and f bark is the frequency variable.

10. A double-edge cochlear dead zone rapid detection system according to claim 1, characterized in that, In the result output module, the preset convergence condition is to satisfy any one of the following conditions: The first one: The difference between the current test result and the previous detection result is within 2 dB for 3 consecutive times or more. The second one: The number of sampling points reaches 64.

Images