A noise environment testing method and system for pure-tone audiometers

Abstract

The application discloses a noise environment test method and system applied to a pure tone audiometer, and the method comprises the following steps: obtaining an anti-interference noise signal stream through a feedforward microphone embedded in a specific curvature area outside an earmuff; inputting the anti-interference noise signal stream into a noise reduction processing module, combining an original pure tone signal input by the audiometer, and outputting a pure tone reservation type reverse signal; superimposing and synthesizing the pure tone reservation type reverse signal and the original pure tone signal to obtain a high-fidelity synthesized audio signal; performing real-time frequency domain analysis on the high-fidelity synthesized audio signal to generate multi-dimensional quality verification data containing distortion characteristics, response speed and fidelity level; and outputting an environment noise adaptive final hearing threshold result based on the comparison result of the multi-dimensional quality verification data and a preset standard threshold. According to the embodiment of the application, efficient suppression of environmental noise and accurate reservation of pure tone signals can be realized, and the reliability and environmental adaptability of hearing threshold test are improved.

Description

Technical Field

[0001] This invention belongs to the field of audio testing technology, and in particular to a method and system for testing noise environment using a pure-tone audiometer. Background Technology

[0002] Traditional pure-tone audiometers are susceptible to interference from ambient sound waves in noisy environments, leading to inaccurate hearing threshold test results. Existing noise reduction technologies mostly employ passive sound insulation or general active noise reduction solutions, which are difficult to optimize for the narrowband characteristics of pure-tone signals in audiometers, often resulting in pure-tone frequency distortion or response delay. For example, while conventional feedforward noise reduction systems can suppress broadband noise, they disrupt the harmonic structure of the pure-tone signal, affecting test accuracy; and fixed-parameter filtering algorithms cannot adapt to dynamically changing noise environments, leading to unstable noise reduction effects. Furthermore, existing technologies lack real-time quality verification of the output signal, failing to guarantee test reliability in complex acoustic scenarios. Summary of the Invention

[0003] The purpose of this invention is to provide a noise environment testing method and system for pure-tone audiometers, in order to overcome the shortcomings of the prior art, achieve efficient suppression of environmental noise and accurate preservation of pure-tone signals, and improve the reliability and environmental adaptability of hearing threshold testing.

[0004] One embodiment of this application provides a noise environment testing method applied to a pure-tone audiometer, the method comprising:

[0005] A feedforward microphone embedded in a specific curvature area on the outside of the earcup is used to collect ambient noise signals at a preset tilt angle. The signals are then amplified by a preamplifier to obtain an anti-interference noise signal stream.

[0006] The anti-interference noise signal stream is input into the noise reduction processing module. Combined with the original pure tone signal input from the audiometer, an anti-phase sound wave with the opposite phase to the noise is generated through an adaptive filtering algorithm. The adaptive filtering algorithm is constrained by the pure tone spectrum characteristics and outputs a pure tone-preserving anti-phase signal.

[0007] The pure tone-preserving inverted signal is superimposed and synthesized with the original pure tone signal, and then the distortion is suppressed by the speaker driver circuit with regulated power supply to obtain a high-fidelity synthesized audio signal.

[0008] Real-time frequency domain analysis is performed on the high-fidelity synthesized audio signal to detect harmonic distortion, transient response and frequency fidelity parameters, and generate multi-dimensional quality verification data including distortion characteristics, response speed and fidelity level.

[0009] Based on the comparison results between the multi-dimensional quality verification data and the preset standard threshold, the pure tone signal gain and noise reduction intensity are dynamically adjusted to output the final hearing threshold result that is adaptive to environmental noise.

[0010] Optionally, the feedforward microphone, embedded in a specific curvature area on the outer side of the earcup, collects ambient noise signals at a preset tilt angle, and performs signal enhancement processing via a preamplifier to obtain an anti-interference noise signal stream, including:

[0011] A feedforward microphone array is deployed in a preset acoustic sensitive area on the outside of the earcups. The tilt angle of the array units is adjusted to form a directional beam, which focuses on collecting ambient noise signals and outputs a highly directional noise signal.

[0012] A highly directional noise signal is input into a multi-band gain controller, which adaptively enhances the gain of the low-frequency band and the human voice band based on the noise energy distribution, suppresses high-frequency random interference, and generates a spectrum-enhanced noise signal.

[0013] The spectral enhancement noise signal is clock-synchronized and sampled. A finite impulse response filter is used to eliminate the phase shift caused by sampling clock jitter and output a phase-synchronized noise signal stream.

[0014] By truncating the burst pulse amplitude in the phase synchronization noise signal stream using a nonlinear limiter and smoothing the baseline fluctuations using a sliding window mean filter, an anti-interference noise signal stream is finally obtained.

[0015] Optionally, the step of inputting the anti-interference noise signal stream into the noise reduction processing module, combining it with the original pure tone signal input from the audiometer, and generating an anti-phase sound wave with the opposite phase to the noise through an adaptive filtering algorithm, wherein the adaptive filtering algorithm is constrained by the pure tone spectrum characteristics and outputs a pure tone-preserving anti-phase signal, includes:

[0016] The raw pure tone signal input from the audiometer is subjected to a fast Fourier transform to identify the core frequency band range with an energy ratio exceeding 80% and generate a pure tone core frequency band marker.

[0017] The anti-interference noise signal stream is input into the adaptive filter, and the weight coefficients are updated based on the least mean square algorithm. At the same time, the pure tone core frequency band marker is introduced as a constraint condition to prohibit the filter from adjusting the weights in this frequency band, and the pure tone protective weight coefficient set is output.

[0018] By using a pure tone-protected weight set to perform convolution operation on the noise signal, a canceled sound wave with a phase strictly opposite to the noise is generated, thus obtaining the basic anti-phase sound wave signal.

[0019] The residual energy in the pure tone segment of the basic anti-phase acoustic signal caused by the constraint conditions is detected, and in-phase compensation micropulses are injected to finally output a pure tone-preserving anti-phase signal.

[0020] Optionally, the step of superimposing the pure tone-preserving inverted signal with the original pure tone signal and performing distortion suppression processing on the speaker driver circuit powered by a regulated voltage to obtain a high-fidelity synthesized audio signal includes:

[0021] The pure tone-preserving inverted signal and the original pure tone signal are input into a digital delay phase-locked loop. The delay of the inverted signal is finely adjusted until the phase difference between it and the pure tone signal approaches zero, and the phase-aligned synthesized signal is output.

[0022] A low-dropout linear regulator provides a millivolt-level precision supply voltage to the speaker driver circuit, suppressing the interference of power supply ripple on the phase-aligned synthesized signal and generating a regulated drive signal.

[0023] The regulated drive signal is input to the predistortion compensator, and the predistortion signal waveform is dynamically predistorted based on the speaker impedance characteristic curve to cancel the nonlinear effect of electroacoustic conversion and obtain the predistortion correction signal.

[0024] The mechanical response delay is calculated based on the speaker diaphragm mass parameters. A lead differential component is added to the predistortion correction signal to output an inertial compensation signal.

[0025] The high-order harmonic components of the speaker output are detected in real time, and a reverse harmonic injection signal is generated and fed back to the inertial compensation signal to finally obtain a high-fidelity synthesized audio signal with a total harmonic distortion of less than 0.5%.

[0026] Optionally, the real-time frequency domain analysis of the high-fidelity synthesized audio signal, detecting harmonic distortion, transient response, and frequency fidelity parameters, and generating multi-dimensional quality verification data including distortion characteristics, response speed, and fidelity level, includes:

[0027] The high-fidelity synthesized audio signal is synchronously input into three analysis channels. The harmonic distortion analyzer captures the total harmonic distortion rate, the transient response capturer records the signal transition delay time, the frequency fidelity detector measures the frequency deviation error, and outputs a three-channel parallel raw index set.

[0028] Features of the current test frequency band are extracted from the three-channel parallel raw index set, and weight coefficients are assigned according to the pure tone test standard. Among them, the low frequency band strengthens the harmonic distortion weight, the mid frequency band improves the frequency fidelity weight, and the high frequency band focuses on the transient response weight, generating a frequency band weighted index vector.

[0029] The real-time environmental noise energy spectrum is collected as the background basis, and the noise coupling component is subtracted from the frequency band weighted index vector to output the denoising calibration index vector.

[0030] The noise reduction calibration index vector is compared with the preset clinical hearing threshold detection standard library, and the compliance score of each index is calculated by the fuzzy decision algorithm to obtain the three-dimensional clinical compliance matrix.

[0031] By integrating harmonic distortion score, transient response level, and frequency fidelity coefficient from the three-dimensional clinical compliance matrix, multi-dimensional quality verification data containing distortion characteristics, response speed, and fidelity level is generated.

[0032] Optionally, the step of dynamically adjusting the pure tone signal gain and noise reduction intensity based on the comparison results of the multi-dimensional quality verification data and the preset standard threshold, and outputting a final hearing threshold result that adapts to environmental noise, includes:

[0033] If the frequency fidelity in the multi-dimensional quality verification data is lower than the preset standard threshold, the gradient boosting algorithm is activated. For each level decrease in fidelity, the noise reduction intensity compensation coefficient is linearly increased, generating a dynamic noise reduction intensity command.

[0034] When the transient response delay exceeds the preset clinical tolerance, the pure tone signal gain is adjusted in reverse according to the delay duration. The gain is increased by 0.5% for every 1 millisecond increase in delay, and a gain compensation command is output.

[0035] The dynamic noise reduction intensity command and gain compensation command are input into the hybrid controller. The conflict resolution algorithm prioritizes frequency fidelity and generates an environmental adaptive control signal.

[0036] Based on the environmental adaptive control signal, the test parameters are adjusted in real time, and the hearing threshold recalibration is completed within 50 milliseconds, and the final hearing threshold result of environmental noise immunity is output.

[0037] Another embodiment of this application provides a noise environment testing system for a pure-tone audiometer, the system comprising:

[0038] The acquisition module is used to acquire ambient noise signals at a preset tilt angle through a feedforward microphone embedded in a specific curvature area on the outside of the earcup. The signal is then amplified by a preamplifier to obtain an anti-interference noise signal stream.

[0039] The input module is used to input the anti-interference noise signal stream into the noise reduction processing module. Combined with the original pure tone signal input from the audiometer, an anti-phase sound wave with the opposite phase to the noise is generated through an adaptive filtering algorithm. The adaptive filtering algorithm is constrained by the pure tone spectrum characteristics and outputs a pure tone-preserving anti-phase signal.

[0040] The synthesis module is used to superimpose the pure tone-preserving inverted signal with the original pure tone signal and synthesize it. The distortion suppression processing is performed by the speaker driver circuit with regulated power supply to obtain a high-fidelity synthesized audio signal.

[0041] The analysis module is used to perform real-time frequency domain analysis on the high-fidelity synthesized audio signal, detect harmonic distortion, transient response and frequency fidelity parameters, and generate multi-dimensional quality verification data including distortion characteristics, response speed and fidelity level.

[0042] The output module is used to dynamically adjust the pure tone signal gain and noise reduction intensity based on the comparison results between the multi-dimensional quality verification data and the preset standard threshold, and output the final hearing threshold result that is adaptive to environmental noise.

[0043] Another embodiment of this application provides a storage medium storing a computer program, wherein the computer program is configured to execute the method described in any of the preceding claims when running.

[0044] Another embodiment of this application provides an electronic device including a memory and a processor, wherein the memory stores a computer program and the processor is configured to run the computer program to perform the method described in any of the preceding claims.

[0045] Compared with existing technologies, this invention provides a noise environment testing method for pure-tone audiometers. It obtains an anti-interference noise signal stream through a feedforward microphone embedded in a specific curvature area on the outer side of the earcup. This anti-interference noise signal stream is then input into a noise reduction processing module, combined with the original pure-tone signal input from the audiometer, to output a pure-tone-preserving inverted signal. The pure-tone-preserving inverted signal is then superimposed with the original pure-tone signal to synthesize a high-fidelity synthesized audio signal. Real-time frequency domain analysis is performed on the high-fidelity synthesized audio signal to generate multi-dimensional quality verification data, including distortion characteristics, response speed, and fidelity level. Based on the comparison results between the multi-dimensional quality verification data and preset standard thresholds, a final hearing threshold result adaptive to environmental noise is output. This achieves efficient suppression of environmental noise and accurate preservation of pure-tone signals, improving the reliability and environmental adaptability of hearing threshold testing. Attached Figure Description

[0046] Figure 1 A hardware structure block diagram of a computer terminal for a noise environment testing method applied to a pure-tone audiometer, provided in an embodiment of the present invention;

[0047] Figure 2 A flowchart illustrating a noise environment testing method for a pure-tone audiometer, provided in an embodiment of the present invention;

[0048] Figure 3 This is a schematic diagram of a noise environment testing system for a pure-tone audiometer, provided as an embodiment of the present invention. Detailed Implementation

[0049] The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0050] The present invention first provides a noise environment testing method for a pure-tone audiometer. This method can be applied to electronic devices, such as computer terminals, specifically ordinary computers.

[0051] The following detailed explanation uses a computer terminal as an example. Figure 1 This is a hardware structure block diagram of a computer terminal used in a noise environment testing method for a pure-tone audiometer, as provided in an embodiment of the present invention. Figure 1 As shown, the computer device includes a processor, memory, and network interface connected via a system bus, wherein the memory may include non-volatile storage media and internal memory.

[0052] The non-volatile storage medium can store an operating system and a computer program. The computer program includes program instructions that, when executed, cause the processor to perform any noise environment testing method applied to a pure-tone audiometer.

[0053] The processor provides computing and control capabilities, supporting the operation of the entire computer device.

[0054] The internal memory provides an environment for the execution of computer programs in non-volatile storage media. When the computer program is executed by the processor, it enables the processor to perform any noise environment testing method applied to a pure-tone audiometer.

[0055] This network interface is used for network communication, such as sending assigned tasks. Those skilled in the art will understand that... Figure 1 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0056] It should be understood that the processor can be a Central Processing Unit (CPU), but it can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among these, a general-purpose processor can be a microprocessor or any conventional processor.

[0057] See Figure 2 The present invention provides a noise environment testing method for a pure-tone audiometer, which may include the following steps:

[0058] S201 uses a feedforward microphone embedded in a specific curvature area on the outside of the earcup to collect ambient noise signals at a preset tilt angle. The signals are then amplified by a preamplifier to obtain an anti-interference noise signal stream.

[0059] Specifically, a feedforward microphone array can be deployed in the acoustically sensitive area on the outside of the earcups, and the tilt angle of the array units can be adjusted to form a directional beam, which can focus on collecting environmental noise signals and output highly directional noise signals.

[0060] The preset acoustic sensitive zone on the outer side of the earcup refers to the high-probability noise incidence area determined by ergonomic simulation and sound field simulation. It is typically located in an arc-shaped recess with a radius of curvature of 5 to 8 millimeters at the lower rear of the earcup. A linear array of four miniature feedforward microphones is deployed in this area using an embedded mounting method. Each microphone unit is approximately 3 millimeters in diameter, with a sensitivity of -38 dB and a frequency response range covering 20 Hz to 16 kHz. The array unit spacing is precisely designed to be half the wavelength of the sound wave (e.g., 1 kHz noise corresponds to a spacing of 17 cm). The tilt angle (TA) of each microphone unit is adjusted via a mechanical knob or an electric servo mechanism, causing its main axis (MA) to tilt outward at an angle of 30 to 45 degrees. When ambient noise (such as the low-frequency hum of an air conditioner, the mid-frequency sound of keyboard typing, or human conversation) is incident, the array uses a beamforming algorithm to sum the time delays (TDS) of the signals from each microphone. For example, for a noise source on the left, a small delay (e.g., 0.1 ms) is added to the signal from the right microphone, ensuring that the noise signals received by all microphones are superimposed in phase in the digital signal processor (DSP), ultimately forming a directional beam (DB) with a width of approximately 60 degrees. This increases the main lobe gain by more than 6 dB while suppressing lateral interference (such as footsteps behind). The signal-to-noise ratio (SNR) of the high-directivity noise signal (HDNS) output by this process is 12 dB higher than that of a single microphone.

[0061] The stability of the directional beam is ensured through an adaptive null-steering technique. The system monitors changes in the azimuth of environmental noise sources in real time (such as a 20-degree azimuth shift caused by people walking), and dynamically calculates the optimal tilt angle combination using a gradient descent algorithm. For example, when a change in the azimuth of the main noise source is detected, the control circuit drives a miniature stepper motor (SM, with a stepping accuracy of 0.9 degrees) to synchronously adjust the tilt angle of all microphone units, ensuring that the main lobe of the beam is always aligned with the direction of the noise source. Simultaneously, a generalized sidelobe canceller (GSC) structure is employed in the DSP: the array signal is divided into a fixed beamforming (FB) path and a blocking matrix (BM) path. The blocking matrix generates a reference signal (RS) containing interference from directions other than the main lobe. The adaptive filter weights are iteratively updated using a least mean square (LMS) algorithm to subtract residual interference components from the main path. This dual protection mechanism ensures the continuous output of a high directional noise signal with a directional error of less than 3 degrees in dynamic environments.

[0062] To address the issue of high-frequency noise scattering, an acoustic metamaterial overlay (AMO) was applied to the surface of the microphone array. This overlay consists of periodically arranged Helmholtz resonators (HR), with each unit measuring 2 mm, capable of absorbing high-frequency scattered wave energy above 8 kHz (absorption rate >90%). Simultaneously, a temperature sensor (TS) and a humidity sensor (HS) are embedded in the array substrate to compensate for beam shift caused by changes in sound velocity in real time (compensation accuracy ±0.5 degrees). Ultimately, under 95 dB of ambient noise, the system outputs a highly directional noise signal with a directivity index (DI) of 8 dB in the target frequency band (500 Hz to 4 kHz), providing a clean noise base for subsequent processing.

[0063] A highly directional noise signal is input into a multi-band gain controller, which adaptively enhances the gain of the low-frequency band and the human voice band based on the noise energy distribution, suppresses high-frequency random interference, and generates a spectrum-enhanced noise signal.

[0064] The Multiband Gain Controller (MGC) employs a 6-channel Butterworth Filter Bank (BFB) to segment the input high-directional noise signal (HDNS). The filter passbands are set as follows: Low-frequency band A (20 Hz–300 Hz), Mid-low-frequency band B (300 Hz–800 Hz), Vocal core band C (800 Hz–2.5 kHz), Mid-high-frequency band D (2.5 kHz–5 kHz), High-frequency band E (5 kHz–10 kHz), and Ultra-high-frequency band F (10 kHz–16 kHz). The Root Mean Square Energy (RMSE) of each sub-band signal is calculated over one second using the Noise Energy Distribution Analyzer (NEDA). When the energy proportion of low frequency band A exceeds 40% of the total energy (typical air conditioner noise scenario), the gain of this band is automatically increased by 3 dB; when the energy of the human voice core band C increases suddenly within 200 milliseconds (ms) (increase > 10 dB), the gain is instantly increased by 6 dB to enhance speech noise capture.

[0065] High-frequency interference suppression employs an Adaptive Spectral Entropy Threshold (ASET). First, the spectral entropy (SE) values ​​for the high-frequency bands E and F are calculated, with the formula representing the probability distribution uncertainty of the signal power spectrum. When the spectral entropy value exceeds a preset threshold of 0.8 (indicating high randomness), the Dynamic Range Compressor (DRC) is activated, attenuating the sub-band gain at a 2:1 compression ratio (e.g., for every 2 dB increase in input, the output only increases by 1 dB). Simultaneously, a notch filter bank (NFB) is inserted in the frequency bands above 8 kHz. The center frequency is automatically located based on real-time spectral peak values ​​(with a step accuracy of 100 Hz), the bandwidth is set to 1 / 3 octave, and the depth is -15 dB, effectively suppressing narrowband interference such as display howling.

[0066] To eliminate transient distortion caused by abrupt gain changes, a smooth transition mechanism (STM) is designed. When a subband requires gain boost, a 20-millisecond Hanning window function is used for ramp transition to avoid signal jumps. After independent gain adjustment, all subband signals are reconstructed into a full-band signal by a synthesis filter bank (SFB). The output spectral-enhanced noise signal (SENS) from this process has a 4 dB signal-to-noise ratio improvement below 300 Hz, a 0.15 improvement in the articulation index (AI) of the human voice band (1 kHz to 2 kHz), and a 8 dB suppression of random noise above 10 kHz.

[0067] The spectral enhancement noise signal is clock-synchronized and sampled. A finite impulse response filter is used to eliminate the phase shift caused by sampling clock jitter and output a phase-synchronized noise signal stream.

[0068] Clock-synchronous sampling (CSS) is driven by a high-stability temperature-compensated crystal oscillator (TCXO) with a reference frequency of 24.576 MHz and phase noise below -150 dBc / Hz at 1 kHz offset. The analog-to-digital converter (ADC) employs a Σ-Δ architecture with a sampling rate of 48 kHz and a resolution of 24 bits. The sampling clock is synchronized with the system master clock via a phase-locked loop (PLL), with jitter controlled within 50 picoseconds (ps). Spectral enhancement noise (SENS) is passed through an anti-aliasing filter (AAF, cutoff frequency 22 kHz) before being input to the ADC, ensuring no spectral folding during sampling.

[0069] To eliminate phase offset (PO) caused by residual jitter in the sampling clock, a 256th-order Finite Impulse Response (FIR) filter is employed. The filter coefficients are designed based on the Least-Squares Linear Phase (LSLP) criterion, with a constant group delay (GD) of 5.3 milliseconds (ms). The input signal enters the FIR step-by-step through a shift register and undergoes a convolution operation (CO) with the pre-stored coefficient vector. The key innovation lies in the introduction of a Jitter Compensation Factor (JCF): by monitoring the phase error signal of the TCXO, the linear phase term (LPT) in the FIR coefficients is dynamically adjusted. For example, when a +100 picosecond (ps) clock lead is detected, a phase delay factor of 0.15% is automatically added to pull the signal phase back to the reference position.

[0070] The phase calibration effect is monitored in real time by the Cross-Correlation Validation Module (CCVM). The current output signal is cross-correlated (CC) with a buffered signal from 1 millisecond (ms) ago; a peak position offset of less than 10 microseconds (μs) is considered acceptable. The final output Phase-Synchronous Noise Signal Stream (PSNSS) exhibits a phase consistency error of less than 0.05 radians (rad) in the 200 Hz to 5 kHz frequency band, providing a precise time-domain reference for subsequent noise reduction.

[0071] By truncating the burst pulse amplitude in the phase synchronization noise signal stream using a nonlinear limiter and smoothing the baseline fluctuations using a sliding window mean filter, an anti-interference noise signal stream is finally obtained.

[0072] The nonlinear limiter (NLL) employs a soft clipping strategy to handle burst pulses. A dynamic threshold is set: the upper limit is 3 times the long-term root mean square (LTRMS) value of the signal, and the lower limit is 0.5 times. When the instantaneous amplitude of the phase synchronization noise signal stream (PSNSS) exceeds the upper limit, hyperbolic tangent function (HTF) compression is initiated: output = upper limit value × tanh(input / upper limit value), with a compression ratio of 10:1. For example, when a door slamming sound with a sudden amplitude increase of 40 dB is detected, the limiter suppresses the peak value to a reasonable range within 0.2 milliseconds (ms), preventing overload of subsequent circuits.

[0073] Baseline fluctuation smoothing is achieved using a sliding window mean filter (SWMF). The window length is adaptively adjusted: a 200-millisecond (ms) long window (approximately 9600 sampling points) is used during steady-state noise periods to ensure smoothness; when the detected signal change rate exceeds 10 dB / s, it automatically switches to a 50-millisecond (ms) short window to improve response speed. The filter uses a ring buffer (RB) to store window data, updating the accumulated sum (AS) with each new sampling point, and the output value is the accumulated sum divided by the current window length. Outlier Exclusion Logic (OEL) is also included: if a sampling point deviates from the window mean by more than 6 dB, it is temporarily removed from the accumulated sum to prevent impulse interference from affecting the baseline estimation.

[0074] The combined operation of limiting and smoothing is coordinated by the Dynamic Range Monitor (DRM). When the limiter is activated more than 5 times per second, the sliding window length is automatically increased to 300 milliseconds (ms) to enhance the smoothing effect; conversely, when the baseline fluctuation standard deviation is less than 0.5 dB, the window is reduced to 100 milliseconds (ms) to preserve signal details. The final output Anti-Interference Noise Signal Stream (AINSS) meets the following requirements: impulse interference suppression ratio greater than 20 dB, baseline fluctuation controlled within ±1 dB, and signal rise time delay less than 8 milliseconds (ms), fully meeting the real-time requirements of subsequent noise reduction modules.

[0075] This method optimizes the sound wave reception path by utilizing a specific curvature structure on the outer side of the earcup. A feedforward microphone array forms a highly directional beam at a preset tilt angle to accurately capture ambient noise signals. The preamplifier dynamically adjusts its gain based on the noise spectrum characteristics, focusing on enhancing low-frequency and human voice signals while suppressing high-frequency random interference. This ensures the acquired noise signal has a high signal-to-noise ratio and phase consistency, achieving high-fidelity acquisition of ambient noise and providing an accurate noise reference signal for subsequent noise reduction processing. Through directional beamforming and spectral enhancement techniques, target noise is effectively separated from background clutter, avoiding the multipath interference issues common with traditional omnidirectional microphones and significantly improving the representativeness and usability of the noise sample.

[0076] S202, the anti-interference noise signal stream is input into the noise reduction processing module, and combined with the original pure tone signal input from the audiometer, an anti-phase sound wave with the opposite phase to the noise is generated through an adaptive filtering algorithm. The adaptive filtering algorithm is constrained by the pure tone spectrum characteristics and outputs a pure tone-preserving anti-phase signal.

[0077] Specifically, a fast Fourier transform can be performed on the raw pure tone signal input from the audiometer to identify the core frequency band range with an energy ratio exceeding 80% and generate a pure tone core frequency band marker.

[0078] Signal frequency domain decomposition and core frequency band location

[0079] The raw pure-tone signal generated by the audiometer (such as a 1000 Hz test tone) is first input to a Fast Fourier Transform (FFTM) module. This module segments the signal at a sampling rate (SR) of 44.1 kHz, with each frame (FR) consisting of 1024 sample points (SP). The time-domain waveform is then converted into a frequency-domain energy spectrum using a butterfly algorithm. This results in a spectrum containing 512 frequency bins (FB), each corresponding to approximately 43 Hz bandwidth (BW). The system calculates the energy value (EV) of all frequency bins, sorts them in descending order, and accumulates the values. When the accumulated energy reaches 80% of the total energy (e.g., a pure tone has 82% energy within the 1000 Hz ± 50 Hz range), this frequency band is identified as the core frequency band (CFB). The core frequency band boundary values ​​(such as 950 Hz to 1050 Hz) are recorded as band markers (BM) in the format [start frequency F_start, end frequency F_end].

[0080] Dynamic frequency band adaptive mechanism

[0081] For different test frequencies (e.g., 250 Hz low frequency and 8000 Hz high frequency), a dynamic threshold strategy (DTS) is used to identify the core bandwidth. For low-frequency pure tones (<1000 Hz), due to the wider vibration range of the cochlear basilar membrane, the bandwidth is allowed to expand to ±15% (e.g., the CFB of a 250 Hz pure tone is 212.5 Hz to 287.5 Hz); for high-frequency pure tones (>4000 Hz), it is compressed to ±3% (e.g., the CFB of an 8000 Hz pure tone is 7760 Hz to 8240 Hz). Simultaneously, the system monitors the noise infiltration rate (NIR) within the CFB in real time. If the NIR exceeds 5% (e.g., noise accounts for 7% of the energy in the 1000 Hz frequency band), the Band Purification Program (BPP) is activated: the noise peak point (NPP) is scanned within the CFB, and a noise isolation zone (NIZ) is generated by extending 20 Hz outward from the NPP. After removing the NIZ from the original CFB, an adjusted core band marker (ACBM) is generated.

[0082] Tag encapsulation and transmission assurance

[0083] The generated pure tone core frequency band marker (e.g., [950, 1050] Hz) is encapsulated into a Digital Marker Frame (DMF), containing a timestamp (TS), center frequency (CF), bandwidth value (BV), and check code (CC). It is transmitted to the noise reduction module via a double-buffered queue (DBQ): when the current frame marker is stored in buffer A (BA), the noise reduction module reads the previous frame marker from buffer B (BB), ensuring that the frequency band data is updated every 20 milliseconds (ms). Cyclic Redundancy Check (CRC-16) is used during transmission to prevent data misalignment (DM).

[0084] The anti-interference noise signal stream is input into the adaptive filter, and the weight coefficients are updated based on the least mean square algorithm. At the same time, the pure tone core frequency band marker is introduced as a constraint condition to prohibit the filter from adjusting the weights in this frequency band, and the pure tone protective weight coefficient set is output.

[0085] Constrained Adaptive Filtering Architecture

[0086] The Anti-Interference Noise Signal Stream (AINSS) is input to the Multi-channel Adaptive Filter (MAF) as a digital stream. This filter contains 256 taps (TP), each corresponding to a weight coefficient (WC). Weight coefficient updates employ the Least Mean Square Algorithm (LMS), with the core formula: WC_new = WC_old + μ × e(n) × x(n), where μ is the Step Size Factor (SSF, default value 0.002), e(n) is the Error Signal (ES), and x(n) is the Noise Input Vector (NIV). The pure tone core frequency band marker (e.g., [950, 1050] Hz) is converted into a frequency mask (FM): the weight coefficient number (e.g., taps 22 to 24) corresponding to the FFT frequency index k (k=22 corresponds to 950 Hz, k=24 corresponds to 1050 Hz) is marked as locked (LS), and the weight coefficient is forced to remain unchanged (ΔWC=0) during LMS iteration.

[0087] Dynamic control strategy for weight coefficients

[0088] In unconstrained frequency bands (such as 0-950 Hz and 1050-20000 Hz), the LMS algorithm iterates normally: the error e(n) between the noise reference signal x(n) and the desired signal d(n) (i.e., the target noise reduction signal) drives the weight coefficient update, making the filter output y(n) approximate d(n). Within constrained frequency bands, a Historical Weight Inheritance (HWI) mechanism is used: the average weight coefficient of the 10 frames before the marker takes effect (e.g., WC_avg=0.173) is recorded as the fixed weight value (FWV) for that frequency band. When environmental noise changes abruptly (e.g., a sudden 60 dB traffic noise), the system initiates the Emergency Unlock Protocol (EUP): if the noise energy increase within the constrained frequency band exceeds 20 dB / s, the lock is temporarily released for 1 second, allowing limited adjustment of the weight coefficients (limited to ΔWC_max=0.01). After this, the lock is restored and the new weight value is saved.

[0089] Weight coefficient set encapsulation and verification

[0090] The weight coefficient set generated after each frame update (e.g., 256 floating-point numbers) undergoes normalization processing (NP) to ensure that the sum of the absolute values ​​of all coefficients is 1 (Σ|WC|=1). An outlier detector (OD) scans for anomalous transition points (e.g., |ΔWC|>0.1). If found, the median value (MV) of the five adjacent coefficients is used instead. The final output is a pure-tone protected weight set (PPWS) with a confidence tag (CT): CT=High when the standard deviation (SD) of the weight coefficients within the constrained frequency band is <0.001, otherwise CT=Medium.

[0091] By using a pure tone-protected weight set to perform convolution operation on the noise signal, a canceled sound wave with a phase strictly opposite to the noise is generated, thus obtaining the basic anti-phase sound wave signal.

[0092] Phase-precise inverse convolution

[0093] The anti-interference noise signal stream (AINSS) is divided into a sliding window (SW) of length 256, and then convolved with a pure tone protected weighted coefficient set (PPWS) in real-time (RTCO). The convolution formula is: y(n) = Σ [x(nk) × WC(k)], where k = 0 to 255, x is the noise signal, and WC is the weighting coefficient. The result y(n) is the basic anti-phase signal (BAS), whose key characteristic is that it is mirror symmetric (MS) with the original noise in the time domain waveform: if the amplitude of the original noise at time t is +A, then the amplitude of BAS at time t is -A, and the phase difference (PD) is stable within the range of 180°±2° (guaranteed by a phase calibration circuit).

[0094] Delay compensation and phase calibration

[0095] To eliminate filter group delay (GD, approximately 5.8 ms), a digital delay line (DDL) is inserted into the noise acquisition path, with a delay duration equal to the filter processing time (PT). Simultaneously, a phase-locked loop (PLL) is used to dynamically monitor the phase difference between the BAS and the original noise: if PD > 182°, a negative micro-adjuster (NM) is activated to increase DDL by 0.1 ms; if PD < 178°, a positive micro-adjuster (PM) is used to decrease DDL by 0.1 ms. After calibration, the phase accuracy (PA) of the BAS is improved to 180° ± 0.5°.

[0096] Amplitude matching and dynamic range control

[0097] The BAS amplitude is adjusted using an Automatic Gain Controller (AGC) to maintain its peak ratio (PR) to the original noise at 1:1 ± 5%. When the ambient noise dynamic range exceeds the limit (e.g., a sudden >90 dB SPL sound pressure level), a soft limiter (SL) is activated: hyperbolic compression (HC) is applied to the BAS amplitude exceeding a threshold (e.g., ±1.0 volts (V)) to prevent speaker overload (OL). The final output BAS signal-to-noise ratio (SNR) is ≥35 dB.

[0098] The residual energy in the pure tone segment of the basic anti-phase acoustic signal caused by the constraint conditions is detected, and in-phase compensation micropulses are injected to finally output a pure tone-preserving anti-phase signal.

[0099] Residual energy detection and quantification

[0100] The basic anti-phase acoustic wave signal (BAS) and the original pure tone signal (e.g., a 1000 Hz sine wave) are input into a differential spectrum analyzer (DSA). The DSA first performs an FFT on the two signals to obtain the spectra S_BAS(f) and S_Pure(f); then it calculates the residual energy spectrum (RES): RES(f) = |S_Pure(f) - S_BAS(f)|, focusing on scanning the core frequency band of the pure tone (e.g., 950-1050 Hz). If RES(f) exceeds a threshold (e.g., -40 dB) at the center frequency, a pure-tone residual (PR) is determined to exist, and the residual level (RL) is quantized as RL = 10 × log10[Σ(RES(f)^2)] (in dB).

[0101] Micropulse generation and injection strategy

[0102] Compensation Micro-pulse (CMP) is generated based on the residual RL level:

[0103] Waveform design: A single-cycle sine wave (SSW) modulated by Hanning window is used, with a center frequency equal to the pure tone test frequency (e.g., 1000 Hz) and a pulse width (PW) of 2 cycles (e.g., 2 ms at 1000 Hz).

[0104] Amplitude calculation: Peak CMP amplitude A_CMP = k × RL, where k is the compensation coefficient (CC, default 0.05 V / dB). If RL = -30 dB, then A_CMP = 1.5 mV.

[0105] Phase synchronization: The positive zero-crossing point (PZP) of the original pure tone signal is locked by a zero-crossing detector (ZCD). A CMP is injected 0.25 cycles after the PZP (e.g., 250 μs at 1000 Hz) to ensure in-phase (IP) with the pure tone.

[0106] Signal synthesis and quality verification

[0107] Compensating micropulses (CMPs) are superimposed on the basic antiphase acoustic signal (BAS) to generate a pure-tone reserved antiphase signal (PRAS). The effect is verified using a residual suppression rate (RSRT) test: within the pure tone core frequency band, the energy ratio (ER) of the PRAS to the original pure tone must satisfy ER ≤ -50 dB. If this is not met, an iterative compensation loop (ICL) is initiated: CMPs are injected a second time in increments of half the current RL until the target is met or the maximum iteration count (Max Iteration Count, MIC=3) is reached. The final output PRAS has a total harmonic distortion (THD) of <1%.

[0108] The noise reduction module employs constrained adaptive filtering technology, monitoring the core frequency band of the pure-tone signal in real time during the generation of inverted sound waves. A frequency domain marking protection mechanism prevents the filter from adjusting weights in critical frequency bands. This algorithm eliminates environmental noise while fully preserving the spectral characteristics of the pure-tone test signal, avoiding signal distortion caused by traditional noise reduction methods. It represents a breakthrough in solving the problem of noise reduction interfering with the test signal, ensuring the accuracy of pure-tone hearing tests. The spectral constraint mechanism ensures that the inverted sound waves cancel only noise components, maintaining the frequency purity and amplitude stability of the original test signal, providing a distortion-free acoustic environment for clinical hearing threshold assessment.

[0109] S203, the pure tone-preserving inverted signal is superimposed and synthesized with the original pure tone signal, and the distortion suppression processing is performed by the speaker driver circuit with regulated power supply to obtain a high-fidelity synthesized audio signal;

[0110] Specifically, the pure tone-preserving inverted signal and the original pure tone signal can be input into a digital delay phase-locked loop, the delay of the inverted signal can be finely adjusted until the phase difference with the pure tone signal approaches zero, and the phase-aligned synthesized signal can be output.

[0111] The system receives a pure-tone-preserving inverted signal (carrying sound wave information out of phase with ambient noise) from the noise reduction module and the original pure-tone signal directly output from the audiometer (i.e., the pure test tone to be played to the subject). Due to different signal processing paths, there may be a microsecond time delay difference (TDD) when the two signals reach the synthesis node, resulting in phase mismatch. This mismatch weakens the subsequent noise cancellation effect. To solve this problem, the system uses a digital delay-locked loop (DDLL) for precise time-domain alignment. The core of the DDLL consists of three parts: a phase detector (PD), a loop filter (LF), and a voltage-controlled delay line (VCDL). Initially, the two signals are input to the phase detector in parallel. The phase detector employs zero-crossing detection (ZCD) technology to compare the zero-crossing time difference (ZCTD) of two signals at a specific frequency point (usually the pure tone frequency) in real time, and quantizes it as a phase error voltage (PEV). This voltage value directly reflects the amplitude and polarity (leading or lagging) of the current phase difference.

[0112] The phase error voltage (PEV) is then fed into a loop filter (LF). The loop filter typically employs a proportional-integral (PI) structure. The proportional component (proportional gain coefficient KP) responds quickly to changes in phase difference, while the integral component (integral gain coefficient KI) accumulates historical errors to eliminate steady-state residual errors. The control voltage (CV) output from the LF drives a voltage-controlled delay line (VCDL). The VCDL consists of a series of programmable digital delay units (such as configurable logic blocks (CLBs) based on a field-programmable gate array (FPGA), whose delay amount (DA) varies linearly with CV. For example, when an inverted signal is detected to lag behind a pure tone signal, PEV is negative, the CV output from the LF increases, and the DA of the VCDL decreases, thereby shortening the propagation path delay of the inverted signal and accelerating its "catching up" with the pure tone signal. Conversely, if the inverted signal leads, DA is increased to make it "wait." This process is closed-loop and continuous. The system sets a very small phase difference convergence threshold (e.g., ±0.1 radians). When the absolute value of the real-time calculated phase difference (PD) is less than this threshold, the two signals are considered to have achieved phase alignment. At this point, DDLL is locked, and the two signals are directly added at the synthesis node, outputting a phase-aligned composite signal (PACS). This signal preserves the integrity of the original pure tone while embedding precisely inverted noise cancellation components, laying the foundation for subsequent high-quality playback.

[0113] To ensure alignment accuracy adapts to different frequency pure tones (e.g., 125 Hz low-frequency pure tones and 8000 Hz high-frequency pure tones), the DDLL parameters (e.g., KP, KI, convergence threshold) can be dynamically configured based on the center frequency (CF) of the current test pure tone. For example, high-frequency pure tones have shorter periods and are more sensitive to time delays; the system automatically increases the KP gain and tightens the convergence threshold to ±0.05 radians. For low-frequency pure tones, the requirements are appropriately relaxed to ±0.2 radians to accelerate the locking speed. Simultaneously, the system has an anti-disturbance mechanism: when sudden changes in environmental noise cause signal waveform distortion, the phase detector will temporarily freeze and update, resuming only after the signal stabilizes, preventing erroneous adjustments. The entire alignment process is typically completed within several signal cycles (e.g., less than 10 milliseconds), ensuring real-time performance. The aligned PACS signal is then sent to the next processing module.

[0114] A low-dropout linear regulator provides a millivolt-level precision supply voltage to the speaker driver circuit, suppressing the interference of power supply ripple on the phase-aligned synthesized signal and generating a regulated drive signal.

[0115] Phase-aligned synthesized signals (PACS) require power amplification to drive the speakers. However, the power supply quality of the driver circuit directly affects signal fidelity. Ripple and noise present in ordinary power supplies can couple into the audio signal, introducing humming or hissing background noise and severely compromising the cancellation effect. Therefore, the system uses a low dropout regulator (LDO) to provide ultra-clean power to the speaker driver circuit. An LDO is a high-performance voltage regulator chip (such as the Texas Instruments TPS7A4700 series), whose core advantage is an extremely small dropout voltage (VDO) between the input voltage (VIN) and the output voltage (VOUT) (as low as 100 mV), and an extremely high power supply rejection ratio (PSRR, e.g., greater than 80 dB at 1 kHz). This means that even with large fluctuations in the input power supply, the LDO can maintain extreme stability of the output voltage.

[0116] The specific power supply architecture is as follows: The main power supply (such as a lithium battery or adapter) is first connected to the input terminal (VIN pin) of the LDO. The LDO internally contains a precision voltage reference (VREF), an error amplifier (EA), a feedback resistor network (FRN), and a power regulator (PT). The error amplifier continuously compares the actual output voltage (Sampled VOUT) sampled by the feedback network with the internal high-precision reference voltage (such as a 1.2V bandgap reference). When it detects that VOUT deviates from the set value due to load changes or input fluctuations (e.g., set to 5.0V to power the drive circuit), the error amplifier outputs an error signal to adjust the conduction level of the power regulator, quickly compensating for the voltage deviation. The system requires the LDO output voltage accuracy to reach the millivolt-level (MVL), for example, within ±5 mV. This is achieved by using a high-precision reference source (temperature drift less than 1 / 10 millionth of a degree Celsius ppm / °C) and a low-temperature drift feedback resistor (e.g., 1 / 5 millionth of a degree Celsius ppm / °C). Simultaneously, a large-capacity, low-equivalent-series-resistance ceramic capacitor (e.g., 22 μF, X7R material, ESR < 5 mΩ) and a low-noise ceramic decoupling capacitor (e.g., 0.1 μF) are connected in parallel at the LDO output to further filter out high-frequency noise and suppress voltage fluctuations caused by transient load changes.

[0117] The clean DC voltage (Stable VOUT) after LDO regulation is directly supplied to the speaker driver circuit (typically a Class AB or Class D audio power amplifier chip). At this point, the Phase Aligned Synthesized Signal (PACS) is fed into the driver circuit as the input signal. Due to the extremely stable supply voltage (ripple peak-to-peak value Vpp less than 1 mV), interference introduced by the power supply is suppressed to a negligible level during signal amplification (e.g., signal-to-noise ratio SNR is increased to greater than 110 dB). The output of the driver circuit is a power-amplified, highly suppressed regulated drive signal (RDS). This signal retains the phase alignment characteristics of the PACS while avoiding additional distortion introduced by power supply noise, providing a "clean" energy source for subsequent speaker electroacoustic conversion.

[0118] The regulated drive signal is input to the predistortion compensator, and the predistortion signal waveform is dynamically predistorted based on the speaker impedance characteristic curve to cancel the nonlinear effect of electroacoustic conversion and obtain the predistortion correction signal.

[0119] As an electroacoustic transducer, a loudspeaker's impedance is not constant but a complex function that varies with frequency, forming the speaker impedance curve (SIC). Simultaneously, loudspeakers also exhibit nonlinear effects such as magnetic circuit nonlinearity and suspension system nonlinearity. This results in a non-ideal linear relationship between the input voltage signal (regulated drive signal RDS) and the output sound pressure level signal, especially at higher volumes or specific frequencies, leading to harmonic distortion and intermodulation distortion. To cancel these distortions at their source, a predistortion compensator (PDC) is introduced after the drive circuit in the system.

[0120] At the core of the predistortion compensator is a real-time digital signal processor (DSP) module, which internally stores the precise impedance characteristic curve (SIC) of the currently used in-ear speaker and a pre-calibrated nonlinear distortion model (NDM). The NDM may contain mathematical relationships (e.g., polynomial fitting or lookup tables) describing the harmonic distortion components (e.g., second harmonic HD2, third harmonic HD3) and intermodulation distortion components (e.g., two-tone difference frequency IMD) of the speaker at different frequencies and drive levels. When the regulated drive signal (RDS) is input to the predistortion compensator, the DSP first analyzes the instantaneous frequency components (IFC) and instantaneous amplitude (IA) of the RDS in real time. Then, it combines the current frequency of the RDS, looks up the SIC to obtain the nominal impedance (Znom) at that frequency, and calculates the current drive current (IDC) based on the IA. More importantly, it queries the nonlinear distortion model (NDM) based on IFC, IA, and Znom to predict which distortion components will be generated if the RDS is directly driven to the loudspeaker (e.g., predicting a second harmonic of -40 dB at 1 kHz and 1 V drive).

[0121] Based on the predicted distortion components, the predistortion compensator dynamically generates a predistortion signal (PDS) with the opposite position and equal amplitude to the predicted distortion. This PDS is intentionally superimposed onto the original regulated drive signal (RDS). For example, if positive-phase second harmonic distortion is predicted, the predistortion compensator injects an anti-phase second harmonic component into the original signal; if intermodulation distortion is predicted, a corresponding difference frequency cancellation component is injected. This process is called waveform predistortion. The signal after superposition, i.e., the predistortion corrected signal (PCS), no longer has the ideal waveform of the original signal, but has been "intentionally distorted". However, when this pre-distorted PCS signal is input to the loudspeaker for electroacoustic conversion, the loudspeaker's own nonlinear distortion effect (such as generating positive second harmonics) will cancel out the pre-set negative second harmonics in the signal. Ultimately, the sound waveform actually emitted by the loudspeaker is closer to the ideal original pure tone-preserving negative signal waveform, significantly reducing total harmonic distortion (such as reducing it to below 0.8%).

[0122] The mechanical response delay is calculated based on the speaker diaphragm mass parameters. A lead differential component is added to the predistortion correction signal to output an inertial compensation signal.

[0123] A loudspeaker's diaphragm has a physical mass (M), which results in mechanical inertia when responding to electrical signals. When the electrical signal changes rapidly (such as a pure tone signal suddenly starting or stopping), the diaphragm cannot instantly reach the target displacement; instead, it requires a certain amount of time to accelerate or decelerate. This manifests as a transient response delay (TRD) or ringing in the acoustic output. This delay is particularly detrimental in hearing tests conducted in noisy environments, as it blurs the start / end points of pure tones and affects the accuracy of hearing threshold judgment. To compensate for this inertia, the system further processes the predistortion correction signal (PCS).

[0124] The system pre-stores the equivalent diaphragm mass parameter (EDMP) of the current earcup speaker unit (typically in milligrams or grams). When playing a pure tone at a specific frequency (f) (especially short, sharp impulses or rapidly rising / falling pure tones), the system calculates the relationship between the ideal acceleration (A) of the diaphragm at that frequency and the input voltage (V) based on classical mechanics principles. This leads to the derivation of the theoretical acceleration delay time (TADT ≈ M / (Bl * V), where Bl is the force-to-electricity conversion factor), caused by the mass (M). Simultaneously, by combining the compliance (Cms) and mechanical resistance (Rms) of the speaker suspension system, the mechanical time constant (MTC) of the entire system can be estimated more accurately.

[0125] Based on the calculated mechanical response delay (e.g., MTC of 0.5 ms), the predistortion compensator (DSP module) performs differentiation on the predistortion correction signal (PCS). Specifically, it calculates the instantaneous rate of change (i.e., the derivative, dV / dt) of the PCS signal. Then, it multiplies this differentiated signal by a lead gain coefficient (LCC) related to the diaphragm quality and the current frequency. The value of the LGC is determined through calibration experiments to ensure that the added component precisely cancels out the inertial delay. Finally, the differentiated signal multiplied by the LGC (i.e., the lead differential component) is superimposed onto the original PCS signal. The resulting signal is called the inertia-compensated signal (ICS). Its effect is that at the instant the original signal is about to rise, the lead differential component generates a spike, prematurely "pushing" the diaphragm to begin moving; at the instant the signal is about to fall, it generates a reverse spike, prematurely "pulling" the diaphragm to decelerate. In this way, the sound actually emitted by the loudspeaker can follow the changes in the electrical signal more quickly and accurately in terms of transient characteristics, significantly improving the rise time and overshoot indicators, and outputting clearer and more accurately positioned pure tone stimulation.

[0126] The high-order harmonic components of the speaker output are detected in real time, and a reverse harmonic injection signal is generated and fed back to the inertial compensation signal to finally obtain a high-fidelity synthesized audio signal with a total harmonic distortion of less than 0.5%.

[0127] Despite pre-distortion compensation and inertia compensation, residual distortion may still occur in actual operation due to temperature changes, aging, or driving conditions exceeding the model range. To achieve the stringent requirement of less than 0.5% total harmonic distortion, the system employs final closed-loop feedback correction. Inside the earcups, close to the speaker diaphragm, is a high-precision, wideband monitoring microphone (MMic). This microphone acquires the actual acoustic output (AAO) signal emitted by the speaker in real time. This AAO signal contains all the acoustic information of the speaker's final output, as well as any residual distortion introduced by the speaker itself.

[0128] The AAO signal acquired by the monitoring microphone is fed into a High-Order Harmonic Analyzer (HOHA). A HOHA typically consists of a high-performance ADC (Analog-to-Digital Converter) and a Fast Fourier Transform (FFT) processor. It performs real-time spectrum analysis (RTSA) on the AAO signal, focusing on extracting high-order harmonic components (HOHC) beyond the fundamental frequency (i.e., the current test pure audio frequency f0), particularly the second harmonic (2f0), the third harmonic (3f0), and higher odd and even harmonics (such as 4f0, 5f0). The HOHA accurately measures the amplitude (AH) and phase (PH) information of these harmonics.

[0129] Based on the amplitude (AH) and phase (PH) of each detected harmonic, the harmonic analyzer generates a corresponding Reverse Harmonic Injection Signal (RHIS). Essentially, the RHIS generates a set of sinusoidal components with the same amplitude as the detected distortion harmonics but strictly out of phase by 180 degrees (out of phase). For example, if a second harmonic distortion with amplitude A2 and phase φ2 is detected, the RHIS will contain a sinusoidal component with amplitude A2 and phase φ2 + 180°. This RHIS signal is fed back in real-time to the preceding signal processing path and superimposed on the previously obtained inertial compensation signal (ICS). This superposition point is typically located before or after the pre-distortion compensator. The new signal superimposed with the RHIS (i.e., the final drive signal) then drives the speaker again. The effect is that the positive-phase harmonic distortion generated by the speaker itself (e.g., +A2*sin(2ωt +φ2)) cancels out the negative-phase harmonic distortion (-A2*sin(2ωt +φ2)) preset in the drive signal at the acoustic output. This is a dynamic closed-loop process: the monitoring microphone continuously listens to the output, the harmonic analyzer continuously detects residual distortion, and generates a new RHIS for fine-tuning compensation. After several iterations (usually within milliseconds), the residual harmonic distortion is greatly suppressed. Ultimately, the total harmonic distortion (THD) of the acoustic signal output by the speaker—that is, the high-fidelity composite audio signal (HFCAS)—is effectively controlled within the preset target (less than 0.5%). This signal provides the subject with pure, stable, and low-distortion pure-tone stimulation in a noisy environment, which is a key guarantee for accurate hearing threshold measurement.

[0130] A digital delay-locked loop (PLL) is used to precisely align the phase of the inverted acoustic wave with the original signal. Millivolt-level voltage regulation circuitry eliminates power supply ripple interference, and loudspeaker pre-distortion compensation technology is incorporated. The system models the nonlinear characteristics of electroacoustic conversion in real time, injecting an inverse correction waveform to effectively suppress harmonic distortion and transient response delay. This achieves end-to-end fidelity of the acoustic signal from digital processing to physical output, controlling the total harmonic distortion of the synthesized audio to below 0.5%. This technology overcomes the shortcomings of traditional audiometers, which are prone to signal distortion in complex noise environments, ensuring the waveform accuracy and timing precision of the tested sound waves.

[0131] S204, perform real-time frequency domain analysis on the high-fidelity synthesized audio signal, detect harmonic distortion, transient response and frequency fidelity parameters, and generate multi-dimensional quality verification data including distortion characteristics, response speed and fidelity level;

[0132] Specifically, a high-fidelity synthesized audio signal can be synchronously input into three analysis channels. The harmonic distortion analyzer captures the total harmonic distortion rate, the transient response capturer records the signal transition delay time, the frequency fidelity detector measures the frequency deviation error, and outputs a three-channel parallel raw index set.

[0133] Signal splitting and dedicated analysis channel construction:

[0134] The high-fidelity composite audio signal (HFCAS) is first input to a high-performance analog-to-digital converter (ADC) for digitization at a sampling rate (SR) much higher than the Nyquist frequency (typically ≥96 kHz), ensuring no loss of signal detail. The digitized signal is then simultaneously copied to three independent digital signal processing (DSP) channels, each dedicated to the real-time analysis of a core parameter:

[0135] Channel 1: Harmonic Distortion Analyzer (HDA). At the heart of this channel is a high-performance digital bandpass filter bank (DBFB), whose center frequency is precisely locked to the current test tone frequency (TTF). After extracting the fundamental wave (FW) signal, the HDA uses a Fast Fourier Transform (FFT) to calculate the energy of its higher harmonics (typically up to the 5th or 7th harmonic). The Total Harmonic Distortion Rate (THDR) is calculated using the following formula (the process is described here): take the square root of the sum of the squares of the harmonic energies (Harmonic Energy, HE), divide by the fundamental energy (Fundamental Energy, FE), and finally multiply by 100% to obtain the percentage value. For example, for a 1 kHz pure tone, HDA will calculate the energy of harmonic components such as 2 kHz, 3 kHz, 4 kHz, and 5 kHz, and combine them to obtain the THDR value (e.g., 0.3%).

[0136] Channel 2: Transient Response Capturer (TRC). This channel is designed to capture the dynamic characteristics of rapidly changing signals. When the audiometer control unit issues a command to switch the amplitude of the pure tone signal (e.g., suddenly increasing from 0 dB to the test intensity), the TRC triggers a high-speed comparator (HSC) to accurately record the time difference that the output signal takes from the initial threshold (e.g., 10% of the target amplitude) to the stable threshold (e.g., 90% of the target amplitude), which is the signal transition delay time (TDT), measured in milliseconds (ms). For example, in a single loudness switch, the TDT might measure 0.8 ms.

[0137] Channel 3: Frequency Fidelity Detector (FFD). The core of this channel is a high-precision Digital Phase-Locked Loop (DPLL). It compares the phase and frequency of the input HFCAS with a reference tone signal (RTS) generated internally by the audiometer, with a frequency accuracy of ±0.1 Hz. By calculating and integrating the instantaneous phase difference (IPD), the frequency deviation error (FDE) is obtained, expressed in Hertz (Hz) or as a percentage (e.g., ±0.05%). For example, if the actual output of a target 1 kHz pure tone is 1000.2 Hz, then the FDE is +0.2 Hz.

[0138] The three channels operate in parallel without interfering with each other, performing hundreds of analyses per second. Ultimately, each channel continuously outputs its raw measurements (THDR, TDT, FDE), forming a Tri-Channel Parallel Raw Indicator Set (TCPRIS).

[0139] Real-time processing and anti-interference guarantee:

[0140] Each analysis channel has independent preprocessing and anti-interference mechanisms:

[0141] Before the FFT, the HDA channel applies a Hanning window function to reduce spectral leakage and employs oversampling (OS) technology to improve low-frequency harmonic resolution. To prevent environmental noise from interfering with harmonic measurements, a narrowband notch filter (NF) is integrated into the channel to specifically suppress known power supply interference frequencies (such as 50 / 60 Hz and its harmonics).

[0142] The TRC channel uses high-speed sampling (e.g., 1 MHz) to capture signal edge details. To avoid false triggering, a Dynamic Threshold Adjustment (DTA) function is implemented: when background noise increases, the absolute values ​​of the start and settling thresholds are automatically increased. The channel also includes a Baseline Stabilizer (BS) to eliminate the influence of DC offset on transition time measurements.

[0143] The FFD channel's DPLL design features a wide capture range (WCR) and a narrow loop bandwidth (NLB). WCR ensures rapid signal lock-in even with large initial frequency offsets; NLB (e.g., a bandwidth of only 1 Hz) effectively filters out out-of-band noise and modulation interference, guaranteeing high accuracy and stability in FDE measurements. The channel incorporates built-in automatic gain control (AGC) to maintain a constant input signal amplitude, preventing spurious frequency offset detection by the DPLL caused by amplitude variations.

[0144] All raw data (THDR, TDT, FDE) undergo short-term smoothing (e.g., 10 ms window) through a moving average filter (MAF) before being output to the next stage to suppress random fluctuations and improve data reliability.

[0145] Data integration and output:

[0146] The Tri-Channel Parallel Raw Metrics Set (TCPRIS) is a structured data stream containing timestamps (TS), the current test frequency (CTF), and the raw measurements from the three channels.

[0147] THDR_value: Total harmonic distortion rate updated in real time (in %).

[0148] TDT_value: The delay time of the most recent signal transition (in ms).

[0149] FDE_value: Real-time frequency offset error (unit: Hz).

[0150] The dataset is packaged and transmitted to the next processing stage at fixed time intervals (e.g., every 5 ms) or event-triggered (e.g., after each amplitude switch). The data format typically employs a lightweight protocol (e.g., a custom binary block or TLV structure) to ensure low latency and high throughput, meeting real-time requirements.

[0151] Features of the current test frequency band are extracted from the three-channel parallel raw index set, and weight coefficients are assigned according to the pure tone test standard. Among them, the low frequency band strengthens the harmonic distortion weight, the mid frequency band improves the frequency fidelity weight, and the high frequency band focuses on the transient response weight, generating a frequency band weighted index vector.

[0152] Frequency band identification and feature extraction:

[0153] After receiving the three-channel parallel raw index set (TCPRIS), the system first extracts key information: the current test frequency (CTF). Based on international hearing testing standards (such as ISO 389), the audible frequency range (typically 125 Hz to 8 kHz) is divided into three main test frequency bands:

[0154] Low-Frequency Band (LFB): 125 Hz - 500 Hz. The sound energy in this band is concentrated and easily interfered with by low-frequency ambient noise (such as the hum of air conditioners and traffic noise). Speakers in this region are also prone to producing greater harmonic distortion (especially the second harmonic).

[0155] Mid-Frequency Band (MFB): 501 Hz - 2000 Hz. This is the most important region for speech intelligibility and hearing threshold sensitivity; frequency fidelity directly affects the accuracy of pitch perception of the test tone.

[0156] High-Frequency Band (HFB): 2001 Hz - 8000 Hz. In this band, the sound wavelength is relatively short, and transient characteristics (such as attack time) are crucial to the clarity and localization of the sound. It is also easily affected by system response delay.

[0157] The frequency band to which the current test belongs is automatically determined based on the CTF value (e.g., if CTF = 1000 Hz, it belongs to MFB). Then, the most relevant raw indicator feature value (FV) to that frequency band is extracted from TCPRIS.

[0158] Standard-based dynamic weight allocation:

[0159] Based on clinical audiological practice and electroacoustic equipment standards (such as IEC 60645), the system presets base weight coefficients (BWC) for the three core indicators in different frequency bands. These weights reflect the importance of each indicator to the final hearing threshold test results in that frequency band:

[0160] Low frequency band (LFB): THDR_weight = 0.6 (enhanced harmonic distortion weight - low frequency distortion easily masks the test tone); FDE_weight = 0.2 (appropriate frequency fidelity weight); TDT_weight = 0.2 (appropriate transient response weight).

[0161] Mid-frequency band (MFB): THDR_weight = 0.3 (appropriate harmonic distortion weight); FDE_weight = 0.5 (weight to improve frequency fidelity - key to pitch accuracy); TDT_weight = 0.2 (appropriate transient response weight).

[0162] High Frequency Band (HFB): THDR_weight = 0.2 (appropriate harmonic distortion weight); FDE_weight = 0.3 (appropriate frequency fidelity weight); TDT_weight = 0.5 (emphasis on transient response weight - affecting sound clarity and temporal resolution).

[0163] The total weighting coefficients are always 1.0. Furthermore, the weights are not entirely fixed. The system fine-tunes the weights by referencing the real-time environmental noise energy spectrum (RENES) (from the feedforward microphone array). For example, if strong low-frequency noise is detected during LFB testing, the system may temporarily increase THDR_weight to 0.7 for more stringent distortion monitoring.

[0164] Weighted calculation and vector generation: The raw index values ​​extracted from the current test frequency band are normalized to fall within the [0,1] interval (0 represents the worst, 1 represents the ideal). The normalization references the preset maximum permissible threshold (MPT) (e.g., THDR_Max = 3%, TDT_Max = 2 ms, FDE_Max = 1 Hz).

[0165] Then, the weighted calculation is performed using the weighting coefficients corresponding to the frequency bands:

[0166] Weighted harmonic distortion (W_THDR) = THDR_weight * (1 - (THDR_value / THDR_Max));

[0167] Weighted Transient Response Value (W_TDT) = TDT_weight *(1 - (TDT_value / TDT_Max));

[0168] Weighted frequency fidelity value (W_FDE) = FDE_weight * (1 - (FDE_value / FDE_Max)).

[0169] These three weighted values ​​(W_THDR, W_TDT, W_FDE) together constitute a three-dimensional band-weighted indicator vector (BWIV). For example, in MFB testing, a possible BWIV is [0.85, 0.92, 0.78], representing the weighted performance scores for distortion, response, and fidelity (the closer to 1, the better). This vector comprehensively reflects the system's output signal quality performance in three key dimensions under the current test frequency band and environment, providing a quantitative basis for subsequent calibration and decision-making.

[0170] The real-time environmental noise energy spectrum is collected as the background basis, and the noise coupling component is subtracted from the frequency band weighted index vector to output the denoising calibration index vector.

[0171] Acquisition and modeling of environmental noise energy spectrum:

[0172] The system continuously acquires ambient noise through the aforementioned feedforward microphone array (FMA). These raw noise signals are fed into a real-time spectrum analysis engine (RSAE). The RSAE uses a short-time fourier transform (STFT) to calculate the energy spectral density (ESD) of the current ambient noise with a short time window (e.g., 50 ms) and an appropriate overlap rate (e.g., 75%), resulting in the real-time ambient noise energy spectrum (RENES). The RENES is a frequency-energy matrix covering the relevant frequency band for hearing tests (e.g., 100 Hz - 10 kHz), with the energy value at each frequency point (Bin) expressed in decibels (dB SPL) or linear amplitude. To facilitate subsequent processing, the system calculates the average energy value (Mean Noise Energy, MNE) of RENES within the critical band (CB) near the current test frequency (CTF) (usually using the equivalent rectangular bandwidth, ERB), which serves as the main background noise floor (BNF) for that test point.

[0173] Modeling and component estimation of noise coupling effects:

[0174] Environmental noise does not exist independently; it couples with the pure tone signal emitted by the audiometer and the electroacoustic characteristics of the system itself, thus affecting the three core metrics measured previously.

[0175] Harmonic distortion (THDR): Strong background noise (especially noise that is in the same frequency as or related to the harmonics of the test tone) may be misidentified as harmonic components by HDA, resulting in an artificially high THDR. The noise coupling component (NCC_THDR) can be modeled as: NCC_THDR ≈ k1 * (MNE / Signal_Level), where k1 is an empirical coefficient and Signal_Level is the sound pressure level (dB SPL) of the current test tone.

[0176] Transient Response (TDT): Sudden noises (such as the sound of a door closing or keyboard typing) may interfere with the TRC's detection of transition edges in pure tone signals, causing TDT measurements to become abnormally large or small. Its impact (NCC_TDT) is difficult to quantify precisely and is usually estimated using statistical methods (such as detecting abnormal jumps in TDT values ​​over a short period) combined with the energy of impulse noise in the current RENES.

[0177] Frequency Fidelity (FDE): Narrowband noise close to the test audio frequency can interfere with the DPLL locking process, causing additional frequency jitter. NCC_FDE ≈ k2 * (Noise_Energy_in_DPLL_Bandwidth / Signal_Energy), where k2 is the DPLL loop signal-to-noise ratio sensitivity coefficient.

[0178] The system's built-in Noise Coupling Model (NCM) estimates the NCC values ​​(NCC_THDR, NCC_TDT, NCC_FDE) for the three indicators in real time based on the current RENES (especially the MNE near the CTF and its harmonics), the test sound intensity (Signal_Level) set by the system, and the characteristic parameters of each analysis channel (such as the NLB bandwidth of DPLL and the filter bandwidth of HDA).

[0179] Component subtraction and calibration vector generation:

[0180] The estimated noise coupling component (NCC) is subtracted from the corresponding element of the band-weighted index vector (BWIV) to eliminate the contamination of the measurement results by environmental noise, resulting in a purer index that reflects the system's performance.

[0181] Calibrated harmonic distortion value (Calibrated W_THDR, C_THDR) = W_THDR - α * NCC_THDR;

[0182] Calibrated transient response value (Calibrated W_TDT, C_TDT) = W_TDT - β * NCC_TDT;

[0183] The calibrated frequency fidelity value (Calibrated W_FDE, C_FDE) = W_FDE - γ * NCC_FDE.

[0184] Here, α, β, and γ are calibration factors (CF) between 0 and 1, used to adjust the intensity of the subtraction. Their values ​​are determined through prior laboratory calibration (comparing the measurement differences with and without noise under known noise conditions). The subtraction operation needs to set a lower limit protection (such as a minimum value of 0) to prevent the result from becoming negative.

[0185] Ultimately, these three calibrated values ​​(C_THDR, C_TDT, C_FDE) form the De-noised Calibrated Indicator Vector (DCIV). This vector more accurately represents the actual quality level of the high-fidelity synthesized audio signal output by the system in the three dimensions of harmonic distortion, transient response, and frequency fidelity under the current environmental noise background, providing a more reliable basis for the final quality assessment and clinical conformity judgment. For example, a DCIV of [0.82, 0.89, 0.75] indicates that after deducting the influence of noise, the indicators remain at a good level.

[0186] The noise reduction calibration index vector is compared with the preset clinical hearing threshold detection standard library, and the compliance score of each index is calculated by the fuzzy decision algorithm to obtain the three-dimensional clinical compliance matrix.

[0187] Construction and use of the clinical hearing threshold testing standard library:

[0188] The system internally stores a predefined Clinical Audiometric Standard Library (PCASL). This library was developed by integrating international / national standards (such as the ISO 389 series for audiometer performance requirements), industry specifications (such as ANSI S3.6), and the high standards of the equipment manufacturers themselves. PCASL defines ideal values ​​(IV) and clinically acceptable limits (CAL) for three core quality parameters for different test frequencies (CTF) and test types (TT) (such as air conduction, bone conduction, and masking). For example:

[0189] During the 1 kHz air conduction test:

[0190] THDR_IV = 0.1%, THDR_CAL = 1.0%;

[0191] TDT_IV = 0.5 ms, TDT_CAL = 1.5 ms;

[0192] FDE_IV = 0 Hz, FDE_CAL = 0.5 Hz.

[0193] During the 4 kHz air conduction test:

[0194] THDR_IV = 0.5%, THDR_CAL = 2.0% (slightly higher tolerance for high-frequency distortion);

[0195] TDT_IV = 0.3 ms, TDT_CAL = 1.0 ms (high frequency focuses more on fast response);

[0196] FDE_IV = 0 Hz, FDE_CAL = 1.0 Hz.

[0197] When the system performs the current test, it will automatically retrieve the corresponding IV and CAL triples (IV_THDR, IV_TDT, IV_FDE) and (CAL_THDR, CAL_TDT, CAL_FDE) from PCASL based on CTF and TT.

[0198] Fuzzy decision-making algorithm for compliance scoring:

[0199] Each element (C_THDR, C_TDT, C_FDE) in the Denoising Calibration Index Vector (DCIV) is compared with the retrieved standard limits (IV and CAL). This comparison is not a simple "pass / fail" binary judgment, but rather employs a Fuzzy Decision Algorithm (FDA) to calculate a Conformance Score (CS) varying between 0 and 100%, providing a more nuanced reflection of the quality level.

[0200] The core of fuzzy decision-making is defining the membership function (MF). Taking harmonic distortion (THDR) as an example:

[0201] Define a fuzzy set of "Excellent Quality" (EQ): the membership degree is 1 (100%) when the index value equals IV, and drops to 0 when the index value reaches a "transition point" (TP) between IV and CAL. TP is usually set to (IV + CAL) / 2.

[0202] Define a fuzzy set of "Acceptable Quality" (AQ): the membership degree is 1 (100%) when the index value is equal to TP, drops to 0 when the index value is equal to IV, and also drops to 0 when the index value is equal to CAL.

[0203] Define a fuzzy set of "Unacceptable Quality" (UQ): the membership degree is 1 (100%) when the index value is equal to CAL, and drops to 0 when the index value is less than TP.

[0204] For the current actual index value (such as the original THDR measurement value corresponding to C_THDR), calculate its membership degree (MD) to the three fuzzy sets EQ, AQ, and UQ. The final compliance score (CS) can be defined as: CS_THDR = (MD_EQ * 100) + (MD_AQ * 70) + (MD_UQ * 0).

[0205] Similarly, CS_TDT and CS_FDE are calculated. Thus, if a THDR value is exactly in IV, then CS_THDR = 100%; if in TP, then CS_THDR = 70%; if in CAL, then CS_THDR = 0%; and between IV and TP, it is linearly interpolated to between 70% and 100%. This method provides a more continuous assessment of the indicator's achievement level, better aligning with human cognitive habits.

[0206] Construct a three-dimensional clinical conformity matrix:

[0207] Applying the above fuzzy decision algorithm to the three metrics in DCIV (harmonic distortion, transient response, and frequency fidelity) yields three corresponding compliance scores:

[0208] CS_THDR: Harmonic distortion compliance score (range 0-100%).

[0209] CS_TDT: Transient response compliance score (range 0-100%);

[0210] CS_FDE: Frequency fidelity compliance score (range 0-100%).

[0211] These three scores together form a three-dimensional vector: [CS_THDR, CS_TDT, CS_FDE], called the 3D Clinical Conformance Matrix (3DCCM). Although named a "matrix," in this context it is a vector containing three elements, each representing the degree of conformity of a key quality dimension relative to clinical standards. For example, a 3DCCM might be [95%, 88%, 92%], indicating that under the current testing conditions, the system's output signal performs excellently in terms of distortion, response speed, and frequency accuracy, approaching the ideal standard. This matrix is ​​the core input for generating the final Multi-dimensional Quality Validation Data (MDQVD).

[0212] By integrating harmonic distortion score, transient response level, and frequency fidelity coefficient from the three-dimensional clinical compliance matrix, multi-dimensional quality verification data containing distortion characteristics, response speed, and fidelity level is generated.

[0213] Data integration and hierarchy mapping:

[0214] The received 3D Clinical Conformity Matrix (3DCCM) contains three numerical scores: CS_THDR, CS_TDT, and CS_FDE (all within the range of 0-100%). To generate more intuitive and easily usable multidimensional quality validation data (MDQVD) for subsequent decision engines, the system first maps these continuous scores to more semantic quality levels (QL). Typically, 3-5 levels are defined: Excellent (E): score ≥90%; Good (G): score ≥75% and <90%; Acceptable (A): score ≥60% and <75%; Poor (P): score <60%.

[0215] For example, CS_THDR = 95% maps to Harmonic Distortion Level (HDL) = E; CS_TDT = 82% maps to Transient Response Level (TRL) = G; and CS_FDE = 92% maps to Frequency Fidelity Level (FFL) = E. This classification simplifies data representation and highlights the quality status.

[0216] Generate structured, multi-dimensional quality objects:

[0217] Multidimensional Quality Verification Data (MDQVD) is a structured data object that integrates information from 3DCCM and adds necessary contextual information to form a comprehensive description of the quality of the current test signal. Its core components include:

[0218] Distortion characteristics (DC):

[0219] Distortion Type: Fixed as "Harmonic Distortion".

[0220] Raw Value: THDR_value from TCPRIS (unit: %).

[0221] Calibrated Value: C_THDR (normalized value) from DCIV.

[0222] Conformance Score: CS_THDR (%) from 3DCCM.

[0223] Quality Level: The HDL (E / G / A / P) obtained through mapping.

[0224] Response Speed ​​(RS):

[0225] Response Type: Fixed as "Transient Response".

[0226] Raw Value: TDT_value from TCPRIS (unit: ms).

[0227] Calibrated Value: C_TDT (normalized value) from DCIV.

[0228] Conformance Score: CS_TDT (%) from 3DCCM.

[0229] Quality Level: The mapped TRL (E / G / A / P).

[0230] Fidelity Level (FL):

[0231] Fidelity Type: Fixed as "Frequency Fidelity".

[0232] Raw Value: FDE_value (unit: Hz) from TCPRIS.

[0233] Calibrated Value: C_FDE (normalized value) from DCIV.

[0234] Conformance Score: CS_FDE (%) from 3DCCM.

[0235] Quality Level: The FFL (E / G / A / P) obtained by mapping.

[0236] Metadata: Current Test Frequency (CTF); Current Test Type (TT); Timestamp (TS); Background Noise Floor (BNF) (from RENES).

[0237] Output and Application:

[0238] The generated Multidimensional Quality Validation Data (MDQVD) is output in a standardized data format (such as JSON or XML). It serves two main purposes:

[0239] Real-time decision support: This serves as the core input for "dynamically adjusting the pure tone signal gain and noise reduction intensity." The decision engine (such as gradient boosting and conflict resolution algorithms) will determine whether and how to adjust parameters (e.g., increase noise reduction intensity or improve pure tone gain) based on the quality level and score of the three dimensions in MDQVD. For example, if the FFL is P (poor), then increasing the noise reduction intensity will be triggered first.

[0240] Quality Monitoring and Recording: The MDQVD, along with the final Hearing Threshold Result (FHTR), is stored in the patient file or device log. This provides an audit trail of signal quality during testing, which is crucial for assessing the reliability of test results, troubleshooting device problems, or meeting quality management system (QMS) requirements. If the MDQVD for a test shows a critical indicator grade of P (Poor), the system can mark the test result as "Questionable" or recommend a retest.

[0241] With MDQVD, the system not only ensures the accuracy of hearing threshold output in noisy environments, but also provides transparency and verifiability regarding the signal quality behind the results, significantly improving the testing reliability and clinical value of pure-tone audiometers in complex environments.

[0242] The system employs a three-channel parallel analysis architecture to simultaneously monitor harmonic components, signal transition delays, and frequency offset errors. A frequency-band adaptive weighting algorithm dynamically adjusts the weights of evaluation indicators based on different test frequencies. The system automatically deducts the coupling effect of environmental noise on the measurement results, generating a clinically oriented quality assessment matrix and establishing an objective and quantitative acoustic quality assessment system, providing data reliability assurance for hearing threshold testing. Multi-dimensional parameter analysis can accurately identify system anomalies, avoiding the risk of misdiagnosis due to equipment performance fluctuations and significantly improving the credibility of test results.

[0243] S205, based on the comparison results of the multi-dimensional quality verification data and the preset standard threshold, dynamically adjust the pure tone signal gain and noise reduction intensity, and output the final hearing threshold result that is adaptive to environmental noise.

[0244] Specifically, if the frequency fidelity in the multi-dimensional quality verification data is lower than a preset standard threshold, the gradient boosting algorithm can be activated. For each level of fidelity decrease, the noise reduction intensity compensation coefficient is linearly increased to generate a dynamic noise reduction intensity command.

[0245] Threshold comparison and algorithm triggering mechanism

[0246] The system receives multi-dimensional quality verification data (including harmonic distortion score, transient response level, and frequency fidelity coefficient) generated in real time from the previous stage. Preset standard thresholds are stored in a clinical hearing threshold testing standard library; for example, the acceptable threshold for frequency fidelity is 90 points (out of 100). When the current frequency fidelity coefficient (e.g., 85 points) is detected to be below the threshold, the Gradient Boosting Algorithm (GBA) is triggered. This algorithm takes the frequency fidelity deviation value as input and divides the fidelity into 5 levels (e.g., 90-100 is Level I, 80-89 is Level II, and so on), each level corresponding to an incremental rule for the Noise Reduction Compensation Coefficient (NRCC). For example, when the fidelity is at Level II (80-89 points), the NRCC is calculated linearly by increasing by 0.1 for every 1 point decrease; if it drops to Level III (70-79 points), the increment increases to 0.15 per point.

[0247] Dynamic generation process of compensation coefficient

[0248] The gradient boosting algorithm incorporates a Level-Increment Mapping Table (LIMT) to determine the unit compensation amount corresponding to the current fidelity level by looking up the table. Assuming the current test frequency is 1kHz (mid-frequency band), and the measured fidelity is 83 points (Level II), the unit compensation amount is 0.1 / minute. Since this is below the threshold of 7 points, the algorithm calculates the NRCC increment as 7 × 0.1 = 0.7. This increment is superimposed on the base noise reduction intensity (initialized by the feedforward microphone array) to generate a Dynamic Noise Reduction Intensity Command (DNRIC). For example, if the base intensity is 50%, it increases to 57% after superposition. This process is completed within 10 milliseconds, ensuring real-time performance.

[0249] Secondary calibration of environmental noise coupling

[0250] To avoid pure-tone signal distortion due to excessive noise reduction, the system simultaneously acquires real-time environmental noise energy spectra (RENES). If the current noise energy is concentrated in the low frequency range (such as the hum of an air conditioner), while the test frequency band is in the mid-frequency range, the NRCC is weighted and corrected using a noise-band coupling factor (NBCF). For example, if the impact factor of low-frequency noise on the mid-frequency test is 0.3, the final NRCC is adjusted to 0.7 × 0.3 = 0.21, and the actual noise reduction intensity is only increased to 50.21%. This dynamic weighting mechanism ensures the accuracy of the compensation coefficients and scene adaptability.

[0251] When the transient response delay exceeds the preset clinical tolerance, the pure tone signal gain is adjusted in reverse according to the delay duration. The gain is increased by 0.5% for every 1 millisecond increase in delay, and a gain compensation command is output.

[0252] Quantization detection of transient response delay

[0253] The transient response capture unit continuously monitors the transition delay time (TDT) of the synthesized audio signal. For example, it measures the time required for the speaker output to reach 90% of the target amplitude when a pure tone signal suddenly increases from 0dB to 80dB. The preset Clinical Tolerance Limit (CTL) is set according to the international standard IEC 60645, typically 5 milliseconds (ms). When the measured TDT value is 6ms (exceeding the CTL by 1ms), the system determines it as an out-of-limit event and triggers the gain compensation process.

[0254] Inverse adjustment logic of gain compensation

[0255] The Inverse Proportional Adjustment Rule (IPAR) is employed: for the portion of the delay exceeding the tolerance (ΔT = TDT - CTL), the pure tone signal gain is increased by 0.5% per millisecond. For example, if ΔT = 1ms, the gain is increased by 0.5%; if ΔT = 2ms (e.g., TDT = 7ms), the increase is 1%. The gain adjustment value is calculated based on the Original Pure Tone Amplitude (OPTA). Assuming OPTA is 70dB, a 0.5% increase results in 70.35dB. This compensation value is encapsulated as a Gain Compensation Command (GCC), which includes the target gain value, the effective timestamp, and the applicable frequency band code.

[0256] Cooperative optimization of diaphragm inertia compensation

[0257] To avoid overloading the speaker diaphragm by simply increasing the gain, the system calls a pre-stored Diaphragm Mass-Frequency Response Model (DM-FRM). For example, during high-frequency testing, the diaphragm has significant inertia, and increasing the gain may exacerbate transient distortion. Therefore, GCC will dynamically increase the slope of the signal rise time by incorporating the inertia compensation signal (including the lead derivative component) from the above steps. For example, when the gain is increased by 0.5%, the intensity of the lead derivative component is simultaneously increased by 10%, ensuring that the diaphragm's acceleration capability matches the gain increase and preventing compensation failure.

[0258] The dynamic noise reduction intensity command and gain compensation command are input into the hybrid controller. The conflict resolution algorithm prioritizes frequency fidelity and generates an environmental adaptive control signal.

[0259] Instruction Fusion Architecture of Hybrid Controller

[0260] The Hybrid Controller (HC) receives two types of commands: Dynamic Noise Reduction Intensity Command (DNRIC) and Gain Compensation Command (GCC). Internally, the controller includes a Command Parsing Layer (CPL) and a Priority Arbitration Layer (PAL). The CPL converts DNRIC into noise reduction filter weight adjustment amounts (e.g., NRCC=0.7 corresponds to a 30% increase in the adaptive filter step size) and resolves GCC into coefficients of the digital gain amplifier (e.g., +0.5% gain corresponds to a coefficient of 1.005).

[0261] Priority strategy of conflict resolution algorithm

[0262] When two types of instructions need to be executed simultaneously (e.g., noise reduction requires decreasing the signal amplitude, while gain compensation requires increasing the amplitude), the Conflict Resolution Algorithm (CRA) is activated. This algorithm follows the principle of "frequency fidelity priority":

[0263] Scenario 1: If the frequency fidelity is below the threshold and the transient response is normal, then execute DNRIC in full and pause GCC;

[0264] Scenario 2: If both are abnormal, but the fidelity deviation level is higher (e.g., fidelity level III vs. latency exceeding the limit by 1ms), then DNRIC will be executed at a rate of 70% and GCC at a rate of 30%.

[0265] Scenario 3: If transient delay threatens the accuracy of hearing threshold (e.g., ΔT>3ms), then GCC priority is temporarily allowed, but the gain increase is limited to no more than 1%.

[0266] The decision result is output through a weighted command fusion matrix (WCFM), for example, the output signal = 0.7 × DNRIC + 0.3 × GCC.

[0267] Generation and verification of environmental adaptive control signals

[0268] The hybrid controller encapsulates the fused parameters into an Environment Adaptive Control Signal (EACS), whose data structure includes: a target noise reduction intensity value (e.g., 57%); a pure tone gain compensation value (e.g., +0.5%); and an effective frequency range (e.g., 1kHz-4kHz).

[0269] Before signal output, the accuracy of the adjusted hearing threshold is predicted using a Virtual Threshold Simulator (VTS). For example, a standard 25 dB HL (hearing level) pure tone is injected to simulate the patient's perceived threshold in a noisy environment. If the prediction error is less than ±2 dB, EACS is approved; otherwise, conflict resolution is retried.

[0270] Based on the environmental adaptive control signal, the test parameters are adjusted in real time, and the hearing threshold recalibration is completed within 50 milliseconds, and the final hearing threshold result of environmental noise immunity is output.

[0271] Real-time parameter adjustment for multi-system linkage

[0272] Environmental Adaptive Control (EACS) signals are distributed to three subsystems via a high-speed bus:

[0273] Noise reduction module: Adjust the step size parameter (SSP) of the adaptive filter according to the noise reduction intensity target value in EACS (e.g., 57%), for example, from 0.01 to 0.013;

[0274] Gain control unit: Modifies the amplitude scaling factor (ASF) of the digital signal processor (DSP) based on the gain compensation value (e.g., +0.5%).

[0275] Speaker driver circuit: Synchronously updates the parameters of the predistortion compensator to offset the risk of harmonic distortion caused by gain increase.

[0276] All parameters are loaded within 5 milliseconds and confirmed by being written to the hardware register.

[0277] Closed-loop verification process for hearing threshold recalibration

[0278] After parameter adjustment, the system immediately performs a three-step recalibration:

[0279] Step 1: Replay the current test pure tone (e.g., 1kHz 30dB HL) and use the three analysis channels to acquire the new quality verification data;

[0280] Step 2: If the frequency fidelity is ≥90 and the transient delay is ≤5ms, the calibration is considered successful; otherwise, EACS is generated iteratively based on the new data.

[0281] Step 3: Verify clinical effectiveness using the Patient Response Feedback Loop (PRFL). For example, if the patient correctly identifies the calibrated pure tone twice consecutively, the hearing threshold is confirmed to be effective.

[0282] The entire process is completed within 50 milliseconds, which is consistent with the time window of human hearing persistence (<100 milliseconds).

[0283] Final hearing threshold output of environmental noise immunity

[0284] The final hearing threshold result (FHTR) includes triple noise reduction protection:

[0285] Data layer: Records noise reduction intensity, gain compensation value, and number of calibrations (e.g., "noise reduction 57%, gain +0.5%, calibration 1 time").

[0286] Labeling layer: Add an "Environmental Noise Immune Tag" (ENIT). The label level is determined by the multi-dimensional quality verification data score (e.g., AAA level: all indicators are better than the threshold by 10%).

[0287] Source tracing layer: Stores the original noise spectrum (acquired by feedforward microphone), adjustment process logs, and virtual hearing threshold simulation results for subsequent auditing.

[0288] The results are output to the audiometer host interface and simultaneously uploaded to the medical database via an encryption protocol to ensure compliance with the ISO8253 standard.

[0289] The system optimizes noise reduction intensity and signal gain in real time using a gradient compensation algorithm based on the deviation of quality verification data. A conflict resolution strategy prioritizes frequency fidelity, completing parameter recalibration within 50 milliseconds to ensure the testing process consistently meets clinical standards. This enables intelligent hearing threshold tracking in noisy environments, overcoming the limitations of traditional equipment that requires manual intervention. The dynamic adjustment mechanism ensures test results are unaffected by environmental noise fluctuations, providing stable and reliable hearing assessment data for clinical use, making it particularly suitable for noisy primary healthcare settings.

[0290] As can be seen, an anti-interference noise signal stream is obtained through a feedforward microphone embedded in a specific curvature area on the outside of the earcup. This anti-interference noise signal stream is then input into a noise reduction processing module, combined with the original pure tone signal input from the audiometer, to output a pure tone-preserving inverted signal. The pure tone-preserving inverted signal is then superimposed and synthesized with the original pure tone signal to obtain a high-fidelity synthesized audio signal. Real-time frequency domain analysis is performed on the high-fidelity synthesized audio signal to generate multi-dimensional quality verification data that includes distortion characteristics, response speed, and fidelity level. Based on the comparison results between the multi-dimensional quality verification data and the preset standard threshold, the final hearing threshold result, which is adaptive to environmental noise, is output. This enables efficient suppression of environmental noise and accurate preservation of pure tone signals, improving the reliability and environmental adaptability of hearing threshold testing.

[0291] Another embodiment of the present invention provides a noise environment testing system for a pure-tone audiometer, see [link to documentation]. Figure 3 The system may include:

[0292] The acquisition module 301 is used to acquire ambient noise signals at a preset tilt angle through a feedforward microphone embedded in a specific curvature area on the outside of the earcup, and then perform signal enhancement processing through a preamplifier to obtain an anti-interference noise signal stream.

[0293] The input module 302 is used to input the anti-interference noise signal stream into the noise reduction processing module, and combine it with the original pure tone signal input from the audiometer to generate an anti-phase sound wave with the opposite phase to the noise through an adaptive filtering algorithm. The adaptive filtering algorithm is constrained by the pure tone spectrum characteristics and outputs a pure tone-preserving anti-phase signal.

[0294] The synthesis module 303 is used to superimpose the pure tone-preserving inverted signal with the original pure tone signal and synthesize it. The distortion suppression processing is performed by the speaker driver circuit powered by voltage regulation to obtain a high-fidelity synthesized audio signal.

[0295] Analysis module 304 is used to perform real-time frequency domain analysis on the high-fidelity synthesized audio signal, detect harmonic distortion, transient response and frequency fidelity parameters, and generate multi-dimensional quality verification data including distortion characteristics, response speed and fidelity level.

[0296] The output module 305 is used to dynamically adjust the pure tone signal gain and noise reduction intensity based on the comparison results of the multi-dimensional quality verification data and the preset standard threshold, and output the final hearing threshold result that is adaptive to environmental noise.

[0297] This invention also provides a storage medium storing a computer program, wherein the computer program is configured to execute the steps in any of the above method embodiments when running.

[0298] Specifically, in this embodiment, the storage medium can be configured to store a computer program for performing the following steps:

[0299] S201 uses a feedforward microphone embedded in a specific curvature area on the outside of the earcup to collect ambient noise signals at a preset tilt angle. The signals are then amplified by a preamplifier to obtain an anti-interference noise signal stream.

[0300] S202, the anti-interference noise signal stream is input into the noise reduction processing module, and combined with the original pure tone signal input from the audiometer, an anti-phase sound wave with the opposite phase to the noise is generated through an adaptive filtering algorithm. The adaptive filtering algorithm is constrained by the pure tone spectrum characteristics and outputs a pure tone-preserving anti-phase signal.

[0301] S203, the pure tone-preserving inverted signal is superimposed and synthesized with the original pure tone signal, and the distortion suppression processing is performed by the speaker driver circuit with regulated power supply to obtain a high-fidelity synthesized audio signal;

[0302] S204, perform real-time frequency domain analysis on the high-fidelity synthesized audio signal, detect harmonic distortion, transient response and frequency fidelity parameters, and generate multi-dimensional quality verification data including distortion characteristics, response speed and fidelity level;

[0303] S205, based on the comparison results of the multi-dimensional quality verification data and the preset standard threshold, dynamically adjust the pure tone signal gain and noise reduction intensity, and output the final hearing threshold result that is adaptive to environmental noise.

[0304] This invention also provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to perform the steps in any of the above method embodiments.

[0305] Specifically, the aforementioned electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the aforementioned processor, and the input / output device is connected to the aforementioned processor.

[0306] Specifically, in this embodiment, the processor can be configured to perform the following steps via a computer program:

[0307] S201 uses a feedforward microphone embedded in a specific curvature area on the outside of the earcup to collect ambient noise signals at a preset tilt angle. The signals are then amplified by a preamplifier to obtain an anti-interference noise signal stream.

[0308] S202, the anti-interference noise signal stream is input into the noise reduction processing module, and combined with the original pure tone signal input from the audiometer, an anti-phase sound wave with the opposite phase to the noise is generated through an adaptive filtering algorithm. The adaptive filtering algorithm is constrained by the pure tone spectrum characteristics and outputs a pure tone-preserving anti-phase signal.

[0309] S203, the pure tone-preserving inverted signal is superimposed and synthesized with the original pure tone signal, and the distortion suppression processing is performed by the speaker driver circuit with regulated power supply to obtain a high-fidelity synthesized audio signal;

[0310] S204, perform real-time frequency domain analysis on the high-fidelity synthesized audio signal, detect harmonic distortion, transient response and frequency fidelity parameters, and generate multi-dimensional quality verification data including distortion characteristics, response speed and fidelity level;

[0311] S205, based on the comparison results of the multi-dimensional quality verification data and the preset standard threshold, dynamically adjust the pure tone signal gain and noise reduction intensity, and output the final hearing threshold result that is adaptive to environmental noise.

[0312] The above description, based on the embodiments shown in the figures, details the structure, features, and effects of the present invention. The above description is only a preferred embodiment of the present invention, but the present invention is not limited to the scope of implementation shown in the figures. Any changes made in accordance with the concept of the present invention, or equivalent embodiments modified to have equivalent changes, that do not exceed the spirit covered by the specification and figures, should be within the protection scope of the present invention.

Claims

1. A method for testing in a noise environment applied to a pure tone audiometer, characterized by, The method comprises: The front feed microphone embedded in the specific curvature area outside the earmuff is used to collect environmental noise signals at a preset inclination angle, and the signal is enhanced by a preamplifier to obtain an anti-interference noise signal stream; The anti-interference noise signal stream is input into a noise reduction processing module, combined with the original pure tone signal input by an audiometer, and an adaptive filtering algorithm is used to generate a reverse phase sound wave opposite to the noise phase, and the adaptive filtering algorithm is constrained by the pure tone spectrum characteristics, and an output pure tone reservation type reverse signal is output; The pure tone reservation type reverse signal is superimposed with the original pure tone signal to generate a high-fidelity synthesized audio signal through distortion suppression processing of a speaker driving circuit with stable voltage supply; Real-time frequency domain analysis is performed on the high-fidelity synthesized audio signal to detect harmonic distortion, transient response and frequency fidelity parameters, and multi-dimensional quality verification data containing distortion characteristics, response speed and fidelity level are generated; Based on the comparison result of the multi-dimensional quality verification data and the preset standard threshold, the pure tone signal gain and the noise reduction intensity are dynamically adjusted, and the final hearing threshold result adaptive to the environmental noise is output.

2. The method of claim 1, wherein, The front feed microphone embedded in the specific curvature area outside the earmuff is used to collect environmental noise signals at a preset inclination angle, and the signal is enhanced by a preamplifier to obtain an anti-interference noise signal stream, comprising: A front feed microphone array is arranged in a preset acoustic sensitive area outside the earmuff, the inclination angle of the array unit is adjusted to form a directional beam, the environmental noise signal is focused and collected, and a high-directivity noise signal is output; The high-directivity noise signal is input into a multi-frequency band gain controller, the low-frequency band and the vocal frequency band gain are adaptively improved based on the noise energy distribution, the high-frequency random interference is suppressed, and a spectrum-enhanced noise signal is generated; The spectrum-enhanced noise signal is clock-synchronized sampled, and a finite impulse response filter is used to eliminate the phase offset caused by the sampling clock jitter, and a phase-synchronized noise signal stream is output; The burst pulse amplitude in the phase-synchronized noise signal stream is truncated by a nonlinear limiter, and the sliding window mean filter is used to smooth the baseline fluctuation, and finally an anti-interference noise signal stream is obtained.

3. The method of claim 2, wherein, The anti-interference noise signal stream is input into a noise reduction processing module, combined with the original pure tone signal input by an audiometer, and an adaptive filtering algorithm is used to generate a reverse phase sound wave opposite to the noise phase, and the adaptive filtering algorithm is constrained by the pure tone spectrum characteristics, and an output pure tone reservation type reverse signal is output; The original pure tone signal input by the audiometer is subjected to fast Fourier transform, the core frequency band range with an energy ratio of more than 80% is identified, and a pure tone core frequency band marker is generated; The anti-interference noise signal stream is input into an adaptive filter, the weight coefficient is updated based on the least mean square algorithm, and the pure tone core frequency band marker is introduced as a constraint condition to prohibit the filter from adjusting the weight value in the frequency band, and a pure tone protection type weight coefficient set is output; The pure tone protection type weight coefficient set is used for convolution operation on the noise signal to generate a counteracting sound wave opposite to the noise phase, and a basic reverse phase sound wave signal is obtained; The residual energy of the pure tone frequency band in the basic reverse phase sound wave signal caused by the constraint condition is detected, a same-phase compensation micro-pulse is injected, and finally a pure tone reservation type reverse signal is output.

4. The method of claim 3, wherein, The pure tone preserving anti-phase signal is superimposed with the original pure tone signal, a speaker driving circuit powered by a stabilized voltage is used for distortion suppression processing, and a high-fidelity synthesized audio signal is obtained, comprising: The pure tone preserving anti-phase signal and the original pure tone signal are input into a digital delay phase-locked loop, the anti-phase signal delay is fine-tuned until the phase difference with the pure tone signal approaches zero, and a phase-aligned synthesized signal is output; A low-dropout linear voltage regulator is used to provide a millivolt-level precision power supply voltage to the speaker driving circuit, to suppress the interference of power supply ripple on the phase-aligned synthesized signal, and to generate a stabilized driving signal; The stabilized driving signal is input into a pre-distortion compensator, the signal waveform is dynamically pre-distorted based on the speaker impedance characteristic curve, the electro-acoustic conversion nonlinearity effect is offset, and a pre-distortion correction signal is obtained; The mechanical response delay is calculated according to the speaker diaphragm mass parameters, a leading differential component is added to the pre-distortion correction signal, and an inertia compensation signal is output; Real-time detection of high-order harmonic components output by the speaker generates a reverse harmonic injection signal fed back to the inertia compensation signal, and finally a high-fidelity synthesized audio signal with total harmonic distortion less than 0.5% is obtained.

5. The method of claim 4, wherein, The high-fidelity synthesized audio signal is subjected to real-time frequency domain analysis, harmonic distortion, transient response and frequency fidelity parameters are detected, and multi-dimensional quality verification data containing distortion characteristics, response speed and fidelity level are generated, comprising: The high-fidelity synthesized audio signal is synchronously input into three analysis channels, wherein the harmonic distortion analyzer captures the total harmonic distortion rate, the transient response capture records the signal transition delay time, and the frequency fidelity detector measures the frequency deviation error, and outputs a three-channel parallel original index set; The current test frequency band characteristics are extracted from the three-channel parallel original index set, and the weight coefficients are allocated according to the pure tone test standard, wherein the low frequency band strengthens the harmonic distortion weight, the middle frequency band improves the frequency fidelity weight, and the high frequency band focuses on the transient response weight, and a frequency band weighted index vector is generated; The real-time environmental noise energy spectrum is collected as the background base, and the noise coupling component is deducted from the frequency band weighted index vector, and a denoising calibration index vector is output; The denoising calibration index vector is compared with the preset clinical hearing threshold detection standard library, the compliance scores of each index are calculated by fuzzy decision algorithm, and a three-dimensional clinical compliance matrix is obtained; The harmonic distortion score, transient response level and frequency fidelity coefficient in the three-dimensional clinical compliance matrix are integrated to generate multi-dimensional quality verification data containing distortion characteristics, response speed and fidelity level.

6. The method of claim 5, wherein, Based on the comparison result of the multi-dimensional quality verification data and the preset standard threshold, the pure tone signal gain and the noise reduction intensity are dynamically adjusted, and the final hearing threshold result adaptive to the environmental noise is output, comprising: If the frequency fidelity in the multi-dimensional quality verification data is lower than the preset standard threshold, a gradient boosting algorithm is started, wherein the noise reduction intensity compensation coefficient is linearly increased by one level for each decrease in fidelity, and a dynamic noise reduction intensity instruction is generated; When the transient response delay exceeds the preset clinical tolerance, the pure tone signal gain is adjusted in reverse according to the delay time, and the gain is increased by 0.5% for every 1 millisecond of delay, and a gain compensation instruction is output. The dynamic noise reduction intensity instruction and the gain compensation instruction are input into a mixed controller, a frequency fidelity is preferentially ensured through a conflict resolution algorithm, and an environment adaptive control signal is generated; Real-time adjustment of test parameters is based on the environment adaptive control signal, and threshold recalibration is completed within 50 milliseconds, and an environment noise immune final threshold result is output.

7. A noise environment testing system applied to a pure tone audiometer, characterized in that, The system comprises: The acquisition module is configured to: The acquisition module is configured to: The input module is configured to: The synthesis module is configured to: The analysis module is configured to:

8. The system of claim 7, wherein, The output module is configured to: The acquisition module is configured to: The storage medium stores a computer program, and the computer program is configured to execute the method in any one of claims 1-6 when running. The memory stores a computer program, and the processor is configured to execute the method in any one of claims 1-6 by running the computer program. ​ 9. A storage medium, characterized by ​ 10. An electronic device comprising a memory and a processor, characterized in that, ​

Images