Children voiceprint self-adaptive correction method and system

Abstract

The invention discloses a children voiceprint self-adaptive correction method and a children voiceprint self-adaptive correction system. The method comprises the steps of obtaining a target child voice signal; multi-dimensional voiceprint features are extracted from the target child voice signal, normalization processing is carried out, a high-dimensional voiceprint feature vector is constructed, and the multi-dimensional voiceprint features comprise MFCC, fundamental frequency and formant; the Euclidean distance between the high-dimensional voiceprint feature vector and the reference high-dimensional voiceprint feature vector is calculated, and a judgment threshold value is combined to carry out matching judgment, so that the final child voiceprint is determined; wherein the judgment threshold is dynamically adjusted by using historical matching data to adapt to sound changes, and the reference high-dimensional voiceprint feature vector is a voiceprint template obtained by combining a screened sample with confidence meeting requirements with a moving average method to progressively update. By implementing the method provided by the invention, the template and the threshold can be dynamically adjusted in combination with a self-adaptive correction mechanism, the error rejection rate is reduced on the premise of ensuring the safety, and the robustness and accuracy of the system when facing the child sounding behavior mode change are improved.

Description

Technical Field

[0001] This invention relates to a voiceprint correction method, and more specifically to a child voiceprint adaptive correction method and system. Background Technology

[0002] In real-world home environments, acoustic conditions are complex and variable, and existing voiceprint recognition systems typically lack the ability to dynamically adjust to these specific scenarios. Children's vocal cords are still developing, and their acoustic characteristics, such as pitch and formant positions, change with age, height, and emotional fluctuations. However, most current voiceprint recognition systems use fixed template matching methods, which cannot update the template in a timely manner to reflect this natural growth process as a child's voice changes slowly. Voiceprint templates are usually frozen during the registration phase and cannot be dynamically corrected based on the gradual changes in a child's acoustic characteristics. This means that the system's recognition performance will inevitably decline over time. Therefore, effectively tracking and adapting to the long-term changes in children's voiceprints has become one of the key issues in improving recognition accuracy.

[0003] Furthermore, children often exhibit rich emotional changes and non-standard pronunciation patterns during speech, such as interspersed crying, laughter, or screaming. These changes may cause short-term drifts in voiceprint features, but traditional voiceprint systems are unable to distinguish which changes are "natural growth" within the normal developmental process and which are "ineffective shifts" caused by external factors. The lack of statistical modeling of changes in emotional state and vocalization makes it difficult for the system to determine the nature of current differences, thus increasing the false rejection rate.

[0004] Therefore, it is necessary to design a new method that can combine an adaptive correction mechanism to dynamically adjust the template and threshold based on long-term interaction data, thereby reducing the false rejection rate while ensuring safety, and improving the robustness and accuracy of the system in the face of changes in children's vocal behavior patterns. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method and system for adaptive correction of children's voiceprints.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a child voiceprint adaptive correction method, comprising:

[0007] Acquire the target child's voice signal;

[0008] Multi-dimensional voiceprint features are extracted from the target child's speech signal and normalized to construct a high-dimensional voiceprint feature vector. The multi-dimensional voiceprint features include MFCC, fundamental frequency, and formants.

[0009] The Euclidean distance between the high-dimensional voiceprint feature vector and the reference high-dimensional voiceprint feature vector is calculated and combined with a decision threshold to perform matching judgment, so as to determine the final child's voiceprint; wherein, the decision threshold is dynamically adjusted by using historical matching data to adapt to voice changes, and the reference high-dimensional voiceprint feature vector is a voiceprint template that is progressively updated by using samples with the required confidence level obtained through screening and the moving average method.

[0010] The further technical solution is as follows: the calculation of the Euclidean distance between the high-dimensional voiceprint feature vector and the reference high-dimensional voiceprint feature vector, combined with a decision threshold, is used for matching and judgment to determine the final child's voiceprint, including:

[0011] Calculate the Euclidean distance between the high-dimensional voiceprint feature vector and the reference high-dimensional voiceprint feature vector;

[0012] Compare the Euclidean distance with the decision threshold;

[0013] When the Euclidean distance is not greater than the decision threshold, it is determined that the high-dimensional voiceprint feature vector and the reference high-dimensional voiceprint feature vector are successfully matched to obtain the final child's voiceprint; when the Euclidean distance is greater than the decision threshold, it is determined that the high-dimensional voiceprint feature vector and the reference high-dimensional voiceprint feature vector are not successfully matched.

[0014] The further technical solution is as follows: the decision threshold is dynamically adjusted using historical matching data to adapt to changes in sound, including:

[0015] Analyze the Euclidean distance values ​​of the most recent successful matches, calculate the mean and standard deviation, and determine the normal fluctuation range of children's voice characteristics;

[0016] Based on the normal fluctuation range of children's voice characteristics, the decision threshold is dynamically updated using an initial threshold plus an adjustment coefficient based on the standard deviation, and the decision threshold is increased when environmental noise causes the signal-to-noise ratio to fall below the noise threshold.

[0017] The further technical solution is as follows: the reference high-dimensional voiceprint feature vector is a voiceprint template that is progressively updated using a combination of selected samples with acceptable confidence levels and a moving average method, including:

[0018] When the Euclidean distance is less than a certain percentage of the decision threshold and after additional verification, the corresponding high-dimensional voiceprint feature vector is used as a sample with the required confidence level.

[0019] The moving average method is used to incorporate samples with the required confidence level into the existing voiceprint template according to the set update rate, so as to update the voiceprint template.

[0020] The further technical solution is as follows: Extracting multi-dimensional voiceprint features from the target child's speech signal and performing normalization processing to construct a high-dimensional voiceprint feature vector includes:

[0021] The target child's speech signal is processed by Mel filter bank, logarithmic energy conversion and discrete cosine transform to generate static MFCC features, and first-order and second-order differences are calculated to capture the temporal variation characteristics of the speech, forming a multidimensional voiceprint feature sub-vector.

[0022] The autocorrelation function method combined with center clipping technology is used to extract the fundamental frequency of the target child's speech signal within a specific frequency range. The first three formant frequencies are solved using the linear predictive coding analysis method. These frequencies, along with the normalized statistical values, are added to the multidimensional voiceprint feature subvector to obtain multidimensional voiceprint features.

[0023] The multi-dimensional voiceprint features are normalized by cepstral mean variance to eliminate channel effects and individual volume differences, and then defined as the current input and registered template feature vectors to obtain high-dimensional voiceprint feature vectors.

[0024] The further technical solution is as follows: The autocorrelation function method combined with center clipping technology is used to extract the fundamental frequency of the target child's speech signal within a specific frequency range. The first three formant frequencies are then solved using a linear predictive coding analysis method. These formsant frequencies, along with normalized statistical values, are appended to the multidimensional voiceprint feature sub-vector to obtain multidimensional voiceprint features, including:

[0025] The fundamental frequency of the target child's speech signal is extracted within a set frequency range by using the autocorrelation function method combined with center clipping technology.

[0026] Using the linear predictive coding analysis method, the prediction order is set, and the frequencies of the first three resonant peaks are solved to obtain the resonant peaks;

[0027] The normalized fundamental frequency mean, variance, and center frequency of the formant are determined and added to the multidimensional voiceprint feature subvector to obtain multidimensional voiceprint features.

[0028] This invention also provides a child voiceprint adaptive correction system, comprising:

[0029] Acquisition unit, used to acquire the target child's speech signal;

[0030] The feature extraction unit is used to extract multi-dimensional voiceprint features from the target child's speech signal, and perform normalization processing to construct a high-dimensional voiceprint feature vector. The multi-dimensional voiceprint features include MFCC, fundamental frequency, and formants.

[0031] A matching unit is used to calculate the Euclidean distance between the high-dimensional voiceprint feature vector and the reference high-dimensional voiceprint feature vector, combined with a decision threshold, to make a matching judgment and determine the final child's voiceprint; wherein, the decision threshold is dynamically adjusted using historical matching data to adapt to voice changes, and the reference high-dimensional voiceprint feature vector is a voiceprint template that is progressively updated using a sliding average method with samples that meet the confidence requirements after screening.

[0032] The further technical solution is as follows: the matching unit includes:

[0033] The distance calculation subunit is used to calculate the Euclidean distance between the high-dimensional voiceprint feature vector and the reference high-dimensional voiceprint feature vector.

[0034] A comparison subunit is used to compare the Euclidean distance with the decision threshold;

[0035] A sub-unit is defined to determine that the high-dimensional voiceprint feature vector and the reference high-dimensional voiceprint feature vector are successfully matched when the Euclidean distance is not greater than the decision threshold, so as to obtain the final child's voiceprint; and to determine that the high-dimensional voiceprint feature vector and the reference high-dimensional voiceprint feature vector are not successfully matched when the Euclidean distance is greater than the decision threshold.

[0036] The further technical solution is as follows: the feature extraction unit includes:

[0037] The sub-vector extraction sub-unit is used to generate static MFCC features from the target child's speech signal through Mel filter bank, logarithmic energy conversion and discrete cosine transform, and to calculate the first and second order differences to capture the temporal variation characteristics of the speech, forming a multidimensional voiceprint feature sub-vector.

[0038] The additional subunit is used to extract the fundamental frequency of the target child's speech signal within a specific frequency range using the autocorrelation function method combined with center clipping technology, and to solve the first three formant frequencies using the linear predictive coding analysis method. These frequencies, along with the normalized statistical values, are then added to the multidimensional voiceprint feature subvector to obtain multidimensional voiceprint features.

[0039] The vector processing subunit is used to normalize the multi-dimensional voiceprint features through cepstral mean variance to eliminate channel effects and individual volume differences, and then define them as the current input and registered template feature vectors to obtain high-dimensional voiceprint feature vectors.

[0040] Its further technical solution is as follows: the additional subunit includes:

[0041] The fundamental frequency extraction module is used to extract the fundamental frequency of the target child's speech signal within a set frequency range using the autocorrelation function method combined with center clipping technology.

[0042] The formant extraction module is used to use the linear predictive coding analysis method, set the prediction order, and solve for the frequencies of the first three formants to obtain the formants.

[0043] The feature addition module is used to determine the normalized fundamental frequency mean, variance, and center frequency of the formant, and add them to the multidimensional voiceprint feature subvector to obtain multidimensional voiceprint features.

[0044] The advantages of this invention compared to existing technologies are as follows: This invention acquires the speech signal of a target child and extracts multi-dimensional voiceprint features, including MFCC, fundamental frequency, and formants, from it. These features are then normalized to construct a high-dimensional voiceprint feature vector. The Euclidean distance between this vector and a reference high-dimensional voiceprint feature vector is calculated, and a dynamically adjusted decision threshold is used for matching to determine the final child's voiceprint identity. The decision threshold is automatically adjusted based on historical matching data to adapt to changes in the child's voice over time. Simultaneously, the reference high-dimensional voiceprint feature vector is progressively updated by selecting samples with high confidence and applying a moving average method, ensuring the accuracy and stability of the voiceprint template. This method achieves dynamic adjustment of the template and threshold based on long-term interaction data, effectively reducing the false rejection rate while ensuring security, and significantly improving the system's robustness and recognition accuracy in the face of changes in children's vocal behavior patterns. This allows the system to better adapt to short-term acoustic feature changes such as timbre changes and emotional fluctuations during a child's natural growth process, ensuring reliable recognition performance over a long period.

[0045] The present invention will be further described below with reference to the accompanying drawings and specific embodiments. Attached Figure Description

[0046] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0047] Figure 1 A flowchart illustrating the adaptive voiceprint correction method for children provided in an embodiment of the present invention;

[0048] Figure 2 This is a schematic block diagram of a child voiceprint adaptive correction system provided in an embodiment of the present invention;

[0049] Figure 3 A schematic block diagram of a computer device provided for an embodiment of the present invention. Detailed Implementation

[0050] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0051] It should be understood that, when used in this specification and the appended claims, the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0052] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0053] It should also be further understood that the term "and / or" as used in this specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0054] Please see Figure 1 , Figure 1 This is a flowchart illustrating the adaptive voiceprint correction method for children provided in this embodiment of the invention. The method acquires and processes the speech signal of the target child, extracts multi-dimensional voiceprint features including MFCC, fundamental frequency, and formants, constructs a high-dimensional voiceprint feature vector, and performs matching judgment with a reference template. The decision threshold is dynamically adjusted using historical matching data to adapt to changes in voice. Simultaneously, a confidence screening mechanism combined with a moving average method is used to update the voiceprint template, ensuring that the system can self-optimize based on long-term interaction data. This reduces the false rejection rate while ensuring safety, enhancing robustness and accuracy in the face of changes in children's vocal behavior patterns. This process includes steps such as validating new samples, calculating the distance to existing templates, dynamically adjusting thresholds, and progressively updating templates, collectively achieving the system's high efficiency and reliability.

[0055] Figure 1 This is a schematic flowchart of the adaptive voiceprint correction method for children provided in an embodiment of the present invention. Figure 1 As shown, the method includes the following steps S110 to S130.

[0056] S110, Acquire the target child's voice signal.

[0057] In this embodiment, the target child's voice signal refers to the voice signal of a specific child extracted from a mixed audio environment containing one or more sound sources.

[0058] Microphone arrays can capture and focus sound sources from specific directions, effectively improving the clarity and signal-to-noise ratio of the target child's speech signal. Neural network models can directly learn to separate the voices of different speakers from mixed audio without explicitly modeling sound source features. Analyzing visual cues (such as mouth movements) in videos helps identify the target child's voice, improving speech recognition accuracy. Real-time feedback from parents confirms the child's identity, helping the system to better update and improve voiceprint templates to distinguish different sound sources.

[0059] It can also receive audio streams containing multiple speakers and background noise, segment the audio into short segments through frame-level preprocessing and optimize them to reduce edge effects; then, features are calculated on each frame to detect multiple sound sources or strong background noise, and the speech signal of the target child is enhanced in a targeted manner using a mask tailored to the characteristics of children's speech, thereby effectively improving the signal-to-noise ratio of the target child's speech; finally, a child's speech signal with high purity is output, which is prepared for subsequent voiceprint feature extraction and matching.

[0060] S120. Extract multi-dimensional voiceprint features from the target child's speech signal and perform normalization processing to construct a high-dimensional voiceprint feature vector, wherein the multi-dimensional voiceprint features include MFCC, fundamental frequency, and formants.

[0061] Step S120 involves extracting multi-dimensional voiceprint features from the target child's speech signal and normalizing them to construct a high-dimensional voiceprint feature vector. Here, "multi-dimensional voiceprint features" refers to various acoustic features extracted from the target child's speech signal that can characterize individual differences, including but not limited to MFCC (Mel-frequency cepstral coefficients), fundamental frequency (F0), and formants. These features together constitute a high-dimensional voiceprint feature vector that can be used to identify and distinguish different speakers.

[0062] In this embodiment, multi-dimensional voiceprint features refer to a high-dimensional feature vector that can comprehensively describe the unique voice characteristics of a speaker by integrating multiple acoustic features (such as MFCC, fundamental frequency, formants, etc.).

[0063] A high-dimensional voiceprint feature vector is a comprehensive feature representation formed by combining various voiceprint features (such as MFCC, fundamental frequency, formants, etc.) and performing necessary preprocessing (such as normalization). This vector can comprehensively reflect the voice characteristics of a speaker, thus providing a basis for subsequent matching and recognition.

[0064] In one embodiment, step S120 described above may include steps S121 to S123.

[0065] S121. Static MFCC features are generated from the target child's speech signal through Mel filter bank, logarithmic energy conversion and discrete cosine transform, and first-order and second-order differences are calculated to capture the temporal variation characteristics of the speech, forming a multidimensional voiceprint feature sub-vector.

[0066] In this embodiment, the multidimensional voiceprint feature sub-vector refers to a specific set of features containing static and dynamic information extracted and combined from these acoustic features, which is used to enhance the speech recognition system's ability to distinguish individual voices.

[0067] First, an FFT transform is performed on the enhanced time-domain signal to obtain the power spectrum. Then, the power spectrum is passed through a set of triangular filters distributed according to the Mel scale. For the characteristics of children's voices, 26 filters were selected to cover the frequency range from 0Hz to 8000Hz.

[0068] Next, the logarithm of the energy output of each filter is taken, and a Discrete Cosine Transform (DCT) is applied. This step removes the correlation between the feature dimensions, producing 12-dimensional MFCC coefficients. Combined with an additional logarithmic energy term, a 13-dimensional static eigenvector is formed.

[0069] To capture the temporal dynamics of speech, first-order difference (Delta) and second-order difference (Delta-Delta) were calculated. In this way, the original 13-dimensional static feature vector was expanded into a 39-dimensional feature sub-vector that includes temporal variation information.

[0070] S122. The autocorrelation function method combined with center clipping technology is used to extract the fundamental frequency of the target child's speech signal within a specific frequency range, and the first three formant frequencies are solved using the linear predictive coding analysis method. These frequencies, along with the normalized statistical values, are added to the multidimensional voiceprint feature subvector to obtain multidimensional voiceprint features.

[0071] In one embodiment, step S122 described above may include steps S1221 to S1223.

[0072] S1221. Using the autocorrelation function method combined with center clipping technology, the fundamental frequency of the target child's speech signal is extracted within a set frequency range;

[0073] S1222. Using the linear predictive coding analysis method, set the prediction order and solve for the frequencies of the first three formants to obtain the formants.

[0074] S1223. Determine the normalized fundamental frequency mean, variance, and center frequency of the formant, and append them to the multidimensional voiceprint feature sub-vector to obtain multidimensional voiceprint features.

[0075] Specifically, the autocorrelation function (ACF) method combined with center clipping technique is used to accurately extract the fundamental frequency of children's speech signals within a specific frequency range (e.g., 200Hz to 800Hz). This method is particularly suitable for analyzing children's voices with high fundamental frequencies and rapid changes. Linear predictive coding (LPC) analysis is used, with an appropriate prediction order (e.g., 12th order) to solve for the frequencies of the first three formants. LPC is an effective model capable of accurately estimating the positions of formants, which is crucial for identifying the acoustic characteristics of different individuals. The mean and variance of the fundamental frequency and the center frequencies of the first three formants are determined and standardized, then added to the previously obtained 39-dimensional feature sub-vectors to form the final multi-dimensional voiceprint features.

[0076] S123. After the multi-dimensional voiceprint features are normalized by cepstral mean variance to eliminate channel effects and individual volume differences, they are defined as the current input and registered template feature vectors to obtain high-dimensional voiceprint feature vectors.

[0077] Cepstral mean-variance normalization (CMVN) is applied to all extracted feature vectors to eliminate the effects of channel effects and individual volume differences. This process ensures that the system maintains high recognition performance even under different environmental or device conditions.

[0078] Through the detailed steps described above, this embodiment effectively extracts pure target child speech signals from complex environments and constructs a highly representative and discriminative high-dimensional voiceprint feature vector, providing a solid foundation for subsequent voiceprint recognition. This meticulous approach not only improves recognition accuracy but also enhances the system's robustness and adaptability.

[0079] In this embodiment, the target child's speech signal is subjected to FFT transformation again to obtain the power spectrum. The power spectrum is then passed through a set of Mel-scale triangular filters. The number of filters M is set to 26, covering a frequency range of 0-8000Hz. The conversion formula between Mel frequency and linear frequency f is: The logarithm of the energy output of each filter is taken, and a Discrete Cosine Transform (DCT) is performed to remove correlations across dimensions, resulting in 12-dimensional MFCC coefficients C(m). These are combined with the logarithmic energy term of a frame to form a 13-dimensional static feature vector. To capture the temporal dynamics of speech, the first-order difference (Delta) and second-order difference (Delta-Delta) are calculated. The formula for calculating the first-order difference is: The final combination yields a basic voiceprint feature vector of 39 dimensions: 13 (static) + 13 (first order) + 13 (second order).

[0080] The fundamental frequency was extracted using the autocorrelation function (ACF) method combined with center clipping. For children's speech, the fundamental frequency search range was set to 200Hz~800Hz. Linear predictive coding (LPC) analysis was used, with a prediction order P=12 (for a 16kHz sampling rate), to solve the LPC equation roots to estimate the frequencies of the first three formants (F1, F2, F3). The normalized fundamental frequency mean, variance, and center frequencies of the first three formants were appended to the MFCC vector to construct a high-dimensional voiceprint feature vector.

[0081] The extracted feature vectors are subjected to cepstral mean and variance normalization (CMVN). For feature dimension j, the normalized features... for: ;in This is the mean of that dimension within the current sliding window. The standard deviation is denoted as . This step eliminates the effects of channel effects and varying speech volumes. The processed eigenvector is denoted as . (Current input) and (Registration template).

[0082] In summary, MFCC, fundamental frequency, and formant features are extracted from the child's speech output in the previous stage, and then normalized to generate the current speaker vector. Retrieves the template vector corresponding to the child bound to the current device from storage. The vector distance d between the two is calculated and compared with the current threshold to make a preliminary identity judgment. If d is less than the threshold and the results are consistent for multiple consecutive frames, it is considered that the voice came from the bound child, and the "pass" result is fed back to the upper-layer application logic; otherwise, the "non-target child or uncertain" status is output.

[0083] S130. Calculate the Euclidean distance between the high-dimensional voiceprint feature vector and the reference high-dimensional voiceprint feature vector, and combine it with a decision threshold to perform matching judgment to determine the final child's voiceprint; wherein, the decision threshold is dynamically adjusted using historical matching data to adapt to voice changes, and the reference high-dimensional voiceprint feature vector is a voiceprint template gradually updated using a sliding average method with samples that meet the confidence requirements after screening.

[0084] In this embodiment, the currently acquired high-dimensional voiceprint feature vector is calculated ( ) and reference high-dimensional voiceprint feature vector ( The Euclidean distance between the two voiceprints is used, combined with a dynamically adjusted decision threshold, to determine the final child's voiceprint. This process adapts to voice changes based on historical matching data, and the reference high-dimensional voiceprint feature vector is a voiceprint template that is gradually updated through confidence screening and a moving average method.

[0085] In one embodiment, step S130 described above may include steps S131 to S133.

[0086] S131. Calculate the Euclidean distance between the high-dimensional voiceprint feature vector and the reference high-dimensional voiceprint feature vector.

[0087] For each recognition request, calculate and The Euclidean distance d is calculated using the following formula: , where D is the total dimension of the feature vector.

[0088] S132. Compare the Euclidean distance with the decision threshold.

[0089] The calculated distance d is then compared to a dynamically adjusted decision threshold Tcur. This threshold is not fixed but automatically adjusted based on the system's multi-round historical records to accommodate the natural fluctuations in the child's voice over time. The threshold adjustment mechanism takes into account the mean and standard deviation of the set of correctly confirmed matching distances from the past L times, as well as factors such as ambient noise levels.

[0090] S133. When the Euclidean distance is not greater than the decision threshold, it is determined that the high-dimensional voiceprint feature vector and the reference high-dimensional voiceprint feature vector are successfully matched to obtain the final child's voiceprint; when the Euclidean distance is greater than the decision threshold, it is determined that the high-dimensional voiceprint feature vector and the reference high-dimensional voiceprint feature vector are not successfully matched.

[0091] If the calculated Euclidean distance d is less than or equal to the current decision threshold Tcur, the speech signal in this identification is considered to come from the target child, and the match is successful; otherwise, if d is greater than Tcur, it is considered a mismatch. For speech signals marked as "high confidence samples", they not only need to meet the requirement of d < 0.8Tcur, but also need to undergo additional verification (such as parental confirmation). These samples will be used to update the reference voiceprint template Vref. The moving average method is used to ensure that the template can slowly follow the timbre changes brought about by the development of the child's vocal cords, while avoiding template mutations caused by a single abnormal pronunciation.

[0092] This method effectively improves the accuracy and robustness of children's voiceprint recognition, especially in complex environments with various background noises and different speakers. Furthermore, through adaptive threshold adjustment and progressive updating of the voiceprint template, the method in this embodiment can well adapt to changes in children's voices as they grow, while also reducing the impact of emotional fluctuations and other factors on the recognition results.

[0093] In this embodiment, the decision threshold is dynamically adjusted using historical matching data to adapt to changes in sound, including:

[0094] Analyze the Euclidean distance values ​​of the most recent successful matches, calculate the mean and standard deviation, and determine the normal fluctuation range of children's voice characteristics;

[0095] Based on the normal fluctuation range of children's voice characteristics, the decision threshold is dynamically updated using an initial threshold plus an adjustment coefficient based on the standard deviation, and the decision threshold is increased when environmental noise causes the signal-to-noise ratio to fall below the noise threshold.

[0096] In this embodiment, the system maintains a historical matching buffer of length L, storing the set of the L most recent correct matching distances {d1, d2, ..., dL}. By performing statistical analysis on these distance values, their mean (mud) and standard deviation (sigmad) are calculated to determine the normal fluctuation range of children's voice characteristics.

[0097] The initial decision threshold is set to Tinit. To adapt to changes in children's voices over time, the system dynamically adjusts the current decision threshold Tcur based on historical data: Tcur = Tinit + βσ d Here, β is an adjustment coefficient (e.g., 1.5), used to allow for natural fluctuations within a certain confidence interval. This means that if a child's voice characteristics change (e.g., due to growth, emotional fluctuations, etc.), the system can automatically adjust its judgment criteria to maintain high recognition accuracy.

[0098] When the ambient noise sensor detects a signal-to-noise ratio (SNR) lower than a preset noise threshold (e.g., 10 dB), the system automatically relaxes the decision threshold to 1.2 times the current threshold Tcur. This reduces the possibility of missed detections in noisy environments and ensures system reliability.

[0099] The reference high-dimensional voiceprint feature vector is a voiceprint template that is progressively updated using a combination of selected samples with acceptable confidence levels and a moving average method, including:

[0100] When the Euclidean distance is less than a certain percentage of the decision threshold and after additional verification, the corresponding high-dimensional voiceprint feature vector is used as a sample with the required confidence level.

[0101] The moving average method is used to incorporate samples with the required confidence level into the existing voiceprint template according to the set update rate, so as to update the voiceprint template.

[0102] In this embodiment, the current feature vector Vcur is considered a "high-confidence sample" only when the recognition distance d satisfies d < 0.8Tcur and after additional verification (such as parental confirmation or prolonged interaction). This strict screening mechanism ensures that only samples that are very close to the target child's real voiceprint can be used to update the template, thereby avoiding misleading the system due to abnormal pronunciation (such as screaming, cold voice) or other similar sounds.

[0103] Once high-confidence samples are identified, the system will update the voiceprint template V using a moving average method. ref The updated formula is as follows: Where lambda is the update rate (e.g., 0.05), this strategy ensures that the template can slowly follow the timbre changes brought about by the development of the child's vocal cords, while avoiding template mutation caused by a single abnormal pronunciation (such as screaming or a cold voice). This method ensures that the template can slowly follow the timbre changes brought about by the development of the child's vocal cords, while preventing the risk of template mutation caused by a single abnormal pronunciation.

[0104] Through the above methods, this embodiment not only significantly improves the stability and robustness of children's voiceprint recognition, but also ensures that the algorithm can run in real time on cost-sensitive toy products, while providing good privacy protection and security.

[0105] Unlike traditional fixed-template recognition techniques, the method in this embodiment proposes an adaptive threshold adjustment and progressive template update mechanism based on historical statistics. This not only solves the problem of long-term drift of voiceprint features caused by children's vocal cord development, but also effectively accommodates short-term acoustic feature changes caused by children's daily emotional fluctuations (such as excitement and fatigue), significantly reducing the system's false rejection rate. To meet the real-time and safety requirements of children's smart toys, this technical solution has undergone targeted optimization in algorithm design, such as using 512-point short frame processing and DCT transformation implemented by lookup table method, which greatly reduces computational complexity and makes it compatible with low-cost, low-power embedded chips (such as MCUs commonly used in toys). At the same time, a strict weighted template update strategy (lambda=0.05) ensures the stability of the voiceprint model, prevents malicious recording attacks or accidental similar sounds from misleading the system, and protects the privacy and safety of children.

[0106] The method in this embodiment significantly improves the recognizability of children's voiceprints in multi-source aliasing scenarios. Through multi-source sound analysis and target child voice source enhancement, the method in this embodiment maintains a high recognition success rate even in the presence of television, parental conversations, and environmental noise. It is specifically optimized for the rich high-frequency components of children's voices, reducing the energy encroachment of interfering sound sources on key frequency bands. This effectively improves the stability and robustness of children's voiceprint recognition. The system, through adaptive thresholds and template update mechanisms, can adapt to the slow changes in voiceprints caused by children's natural growth. It is tolerant of emotional fluctuations and changes in vocalization patterns, avoiding frequent false rejections or misjudgments due to one or two unusual vocalizations. It meets the real-time and safety requirements of children's smart toys. The entire algorithm structure was designed with full consideration of the computing power limitations of the toy, enabling real-time operation on low- to mid-range processors. Through a strict template update strategy and parental confirmation mechanism, it ensures recognition convenience while preventing easy impersonation by other children or adults.

[0107] The specific implementation includes three main stages of the recognition process: acquiring the child's speech; secondly, in the voiceprint extraction and matching stage, performing MFCC, fundamental frequency, and formant feature extraction on the child's speech output from the previous stage, and performing normalization processing to generate the current voiceprint vector v. cur Read the template vector v corresponding to the child currently bound to the device from storage. ref The system calculates the vector distance d between the two children and compares it with the current threshold to make a preliminary identity judgment. If d is less than the threshold and the results are consistent across multiple consecutive frames, the system considers the voice to be from the bound child and sends a "pass" result to the upper-layer application logic. Otherwise, it outputs a "non-target child or uncertain" status. Finally, in the scene adaptation and self-adaptation stage, when the system determines that the current recognition environment is stable and multiple recognition results are confirmed by the parent or application side, the corresponding v_cur is marked as a high-confidence sample and stored in the local cache. In outdoor scenes or strong noise environments, the system dynamically improves the multi-source detection sensitivity and adjusts the threshold and mask strategy accordingly to reduce the probability of false positives. In scenes with overlapping voices, if multiple children's voice sources are identified, the system can prioritize the main user's child based on the device binding relationship and historical interaction frequency, and suppress the influence of other children's voice sources on voiceprint matching accordingly.

[0108] The domain division involves multi-source speech separation technology and children's voiceprint recognition technology. Multi-source speech separation technology involves modeling and distinguishing multiple speech signals from parents, children, television, music, etc., within the same space; it involves sound source separation and enhancement algorithms targeting different sound source power spectrum distributions, spatial locations, and temporal structure characteristics; and it involves real-time processing and online adaptation of aliased speech on low-computing-power embedded platforms. Children's voiceprint recognition technology involves building robust acoustic feature extraction models within the range of children's voice frequencies, taking into account the characteristics of concentrated high-frequency energy and rapid frequency fluctuations in children's voices; it involves the design of matching strategies based on multi-dimensional voiceprint vectors, including distance measurement, threshold setting, and adaptive updates; and it involves dynamic correction of children's voiceprint stability, ensuring the system maintains reliable recognition even under conditions of pitch changes, emotional fluctuations, and environmental changes.

[0109] Application scenarios include children's smart toys and multi-terminal collaborative scenarios in the home. It supports accurate identification of the voiceprint of the bound child in multi-sound source scenarios such as the living room and bedroom, enabling personalized addressing and dialogue; when siblings and parents speak simultaneously, it prioritizes and locks onto the target child's voice source to avoid confusion in dialogue; it executes differentiated interaction strategies based on voiceprint results, such as switching story recommendations and learning content difficulty according to the child's identity. When smart TVs, speakers, and toys are present in the home, voiceprint recognition enables "toy-first response" for children; it supports linkage with the parent's mobile app to record voice interaction logs or learning progress based on the identified child's identity; and when multiple devices pick up sound simultaneously, the voiceprint recognition results of the method in this embodiment are used for role differentiation and access control. To ensure the stability of children's voiceprints, the system automatically models the "natural fluctuation range" of children's voiceprints during multiple interactions, allowing the system to tolerate a certain degree of pitch, speech rate, and emotional changes. An adaptive template update mechanism based on historical records ensures that the system can still correctly identify children even when their voices change slowly with age. When extreme tones (such as screaming or crying) are detected, a specific strategy is employed to prevent such atypical speech from being mistakenly recorded in the voiceprint template or triggering incorrect identity determinations. To improve voiceprint matching accuracy, a multi-dimensional voiceprint vector is constructed, including MFCC (Mel-Frequency Cepstral Coefficients), F0 (Fundamental Frequency), formant positions, and other robust time-frequency features, enhancing the ability to distinguish individual differences. Distance metrics and decision thresholds suitable for children's voiceprints are designed to avoid systematic biases caused by direct transfer from adult models. An adaptive correction mechanism dynamically adjusts the template and thresholds based on long-term interaction data, reducing the false rejection rate while ensuring safety.

[0110] The aforementioned adaptive voiceprint correction method for children acquires the speech signal of the target child and extracts multi-dimensional voiceprint features, including MFCC, fundamental frequency, and formants, from it. These features are then normalized to construct a high-dimensional voiceprint feature vector. The Euclidean distance between this vector and a reference high-dimensional voiceprint feature vector is calculated, and a dynamically adjusted decision threshold is used for matching to determine the final child's voiceprint identity. The decision threshold is automatically adjusted based on historical matching data to adapt to changes in children's voices over time. Simultaneously, the reference high-dimensional voiceprint feature vector is progressively updated by selecting high-confidence samples and applying a moving average method, ensuring the accuracy and stability of the voiceprint template. This method dynamically adjusts the template and threshold based on long-term interaction data, effectively reducing the false rejection rate while ensuring safety, and significantly improving the system's robustness and recognition accuracy in the face of changes in children's vocal behavior patterns. This allows the system to better adapt to short-term acoustic feature changes such as timbre changes and emotional fluctuations during a child's natural growth process, ensuring reliable recognition performance over long periods.

[0111] Figure 2 This is a schematic block diagram of a child voiceprint adaptive correction system 300 provided in an embodiment of the present invention. Figure 2 As shown, corresponding to the above-described adaptive voiceprint correction method for children, the present invention also provides a child voiceprint adaptive correction system 300. This child voiceprint adaptive correction system 300 includes a unit for performing the above-described adaptive voiceprint correction method for children, and the system can be configured in a server. Specifically, please refer to... Figure 2 The child voiceprint adaptive correction system 300 includes an acquisition unit 301, a feature extraction unit 302, and a matching unit 303.

[0112] The acquisition unit 301 is used to acquire the target child's speech signal; the feature extraction unit 302 is used to extract multi-dimensional voiceprint features from the target child's speech signal, perform normalization processing, and construct a high-dimensional voiceprint feature vector, wherein the multi-dimensional voiceprint features include MFCC, fundamental frequency, and formants; the matching unit 303 is used to calculate the Euclidean distance between the high-dimensional voiceprint feature vector and the reference high-dimensional voiceprint feature vector, combined with a decision threshold, to perform matching judgment to determine the final child's voiceprint; wherein the decision threshold is dynamically adjusted using historical matching data to adapt to voice changes, and the reference high-dimensional voiceprint feature vector is a voiceprint template gradually updated using a sliding average method with samples that meet the confidence requirements after screening.

[0113] In one embodiment, the matching unit 303 includes:

[0114] The system includes a distance calculation subunit for calculating the Euclidean distance between the high-dimensional voiceprint feature vector and the reference high-dimensional voiceprint feature vector; a comparison subunit for comparing the Euclidean distance with a decision threshold; and a determination subunit for determining that the high-dimensional voiceprint feature vector and the reference high-dimensional voiceprint feature vector are successfully matched when the Euclidean distance is not greater than the decision threshold, thus obtaining the final child's voiceprint; and determining that the high-dimensional voiceprint feature vector and the reference high-dimensional voiceprint feature vector are not successfully matched when the Euclidean distance is greater than the decision threshold.

[0115] In one embodiment, the feature extraction unit 302 includes:

[0116] The sub-vector extraction sub-unit is used to generate static MFCC features from the target child's speech signal through Mel filter bank, logarithmic energy conversion, and discrete cosine transform, and to calculate the first and second order differences to capture the temporal variation characteristics of the speech, forming a multidimensional voiceprint feature sub-vector. The appending sub-unit is used to extract the fundamental frequency of the target child's speech signal within a specific frequency range using the autocorrelation function method combined with center clipping technology, and to solve for the first three formant frequencies using the linear predictive coding analysis method, and append them together with the normalized statistical values ​​to the multidimensional voiceprint feature sub-vector to obtain multidimensional voiceprint features. The vector processing sub-unit is used to normalize the multidimensional voiceprint features through cepstral mean variance to eliminate channel effects and individual volume differences, and then define them as the current input and registered template feature vectors to obtain a high-dimensional voiceprint feature vector.

[0117] In one embodiment, the additional subunit includes:

[0118] The fundamental frequency extraction module is used to extract the fundamental frequency of the target child's speech signal within a set frequency range using the autocorrelation function method combined with center clipping technology; the formant extraction module is used to use the linear predictive coding analysis method, set the prediction order, and solve for the frequencies of the first three formants to obtain the formants; the feature addition module is used to determine the normalized fundamental frequency mean, variance, and center frequency of the formants, and add them to the multidimensional voiceprint feature subvector to obtain multidimensional voiceprint features.

[0119] It should be noted that those skilled in the art can clearly understand that the specific implementation process of the above-mentioned child voiceprint adaptive correction system 300 and each unit can be referred to the corresponding description in the foregoing method embodiments. For the sake of convenience and brevity, it will not be repeated here.

[0120] The aforementioned child voiceprint adaptive correction system 300 can be implemented as a computer program, which can, for example... Figure 3 It runs on the computer device shown.

[0121] Please see Figure 3 , Figure 3 This is a schematic block diagram of a computer device provided in an embodiment of this application. The computer device 500 can be a server, wherein the server can be a standalone server or a server cluster composed of multiple servers.

[0122] See Figure 3 The computer device 500 includes a processor 502, a memory, and a network interface 505 connected via a system bus 501. The memory may include a non-volatile storage medium 503 and internal memory 504.

[0123] The non-volatile storage medium 503 may store an operating system 5031 and a computer program 5032. The computer program 5032 includes program instructions that, when executed, cause the processor 502 to perform a child voiceprint adaptive correction method.

[0124] The processor 502 provides computing and control capabilities to support the operation of the entire computer device 500.

[0125] The internal memory 504 provides an environment for the operation of the computer program 5032 in the non-volatile storage medium 503. When the computer program 5032 is executed by the processor 502, the processor 502 can perform a child voiceprint adaptive correction method.

[0126] This network interface 505 is used for network communication with other devices. Those skilled in the art will understand that... Figure 3 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 500 to which the present application is applied. The specific computer device 500 may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0127] The processor 502 is used to run a computer program 5032 stored in a memory to implement all the steps of the child voiceprint adaptive correction method.

[0128] It should be understood that in the embodiments of this application, the processor 502 may be a central processing unit (CPU), or it may 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. The general-purpose processor may be a microprocessor or any conventional processor.

[0129] It will be understood by those skilled in the art that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program includes program instructions and can be stored in a storage medium, which is a computer-readable storage medium. The program instructions are executed by at least one processor in the computer system to implement the process steps of the embodiments of the above methods.

[0130] Therefore, the present invention also provides a storage medium. This storage medium may be a computer-readable storage medium. The storage medium stores a computer program, wherein when executed by a processor, the computer program causes the processor to perform all steps of the child voiceprint adaptive correction method.

[0131] The storage medium can be any computer-readable storage medium capable of storing program code, such as a USB flash drive, portable hard drive, read-only memory (ROM), magnetic disk, or optical disk.

[0132] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0133] In the embodiments provided by this invention, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the system embodiments described above are merely illustrative. For example, the division of each unit is only a logical functional division, and there may be other division methods in actual implementation. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed.

[0134] The steps in the method of this invention can be adjusted, merged, or reduced in order according to actual needs. The units in the system of this invention can be merged, divided, or reduced according to actual needs. Furthermore, the functional units in the various embodiments of this invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0135] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, a terminal, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention.

[0136] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for adaptive correction of children's voiceprints, characterized in that, include: Acquire the target child's voice signal; Multi-dimensional voiceprint features are extracted from the target child's speech signal and normalized to construct a high-dimensional voiceprint feature vector. The multi-dimensional voiceprint features include MFCC, fundamental frequency, and formants. The Euclidean distance between the high-dimensional voiceprint feature vector and the reference high-dimensional voiceprint feature vector is calculated and combined with a decision threshold to perform matching judgment, so as to determine the final child's voiceprint; wherein, the decision threshold is dynamically adjusted by using historical matching data to adapt to voice changes, and the reference high-dimensional voiceprint feature vector is a voiceprint template that is progressively updated by using samples with the required confidence level obtained through screening and the moving average method.

2. The adaptive voiceprint correction method for children according to claim 1, characterized in that, The calculation of the Euclidean distance between the high-dimensional voiceprint feature vector and the reference high-dimensional voiceprint feature vector, combined with a decision threshold, for matching judgment to determine the final child's voiceprint, includes: Calculate the Euclidean distance between the high-dimensional voiceprint feature vector and the reference high-dimensional voiceprint feature vector; Compare the Euclidean distance with the decision threshold; When the Euclidean distance is not greater than the decision threshold, it is determined that the high-dimensional voiceprint feature vector and the reference high-dimensional voiceprint feature vector are successfully matched to obtain the final child's voiceprint; when the Euclidean distance is greater than the decision threshold, it is determined that the high-dimensional voiceprint feature vector and the reference high-dimensional voiceprint feature vector are not successfully matched.

3. The adaptive voiceprint correction method for children according to claim 1, characterized in that, The decision threshold is dynamically adjusted using historical matching data to adapt to changes in sound, including: Analyze the Euclidean distance values ​​of the most recent successful matches, calculate the mean and standard deviation, and determine the normal fluctuation range of children's voice characteristics; Based on the normal fluctuation range of children's voice characteristics, the decision threshold is dynamically updated using an initial threshold plus an adjustment coefficient based on the standard deviation, and the decision threshold is increased when environmental noise causes the signal-to-noise ratio to fall below the noise threshold.

4. The adaptive voiceprint correction method for children according to claim 1, characterized in that, The reference high-dimensional voiceprint feature vector is a voiceprint template that is progressively updated using a combination of selected samples with acceptable confidence levels and a moving average method, including: When the Euclidean distance is less than a certain percentage of the decision threshold and after additional verification, the corresponding high-dimensional voiceprint feature vector is used as a sample with the required confidence level. The moving average method is used to incorporate samples with the required confidence level into the existing voiceprint template according to the set update rate, so as to update the voiceprint template.

5. The adaptive voiceprint correction method for children according to claim 1, characterized in that, The step of extracting multi-dimensional voiceprint features from the target child's speech signal, performing normalization processing, and constructing a high-dimensional voiceprint feature vector includes: The target child's speech signal is processed by Mel filter bank, logarithmic energy conversion and discrete cosine transform to generate static MFCC features, and first-order and second-order differences are calculated to capture the temporal variation characteristics of the speech, forming a multidimensional voiceprint feature sub-vector. The autocorrelation function method combined with center clipping technology is used to extract the fundamental frequency of the target child's speech signal within a specific frequency range. The first three formant frequencies are solved using the linear predictive coding analysis method. These frequencies, along with the normalized statistical values, are added to the multidimensional voiceprint feature subvector to obtain multidimensional voiceprint features. The multi-dimensional voiceprint features are normalized by cepstral mean variance to eliminate channel effects and individual volume differences, and then defined as the current input and registered template feature vectors to obtain high-dimensional voiceprint feature vectors.

6. The adaptive voiceprint correction method for children according to claim 5, characterized in that, The method employs the autocorrelation function method combined with center clipping technology to extract the fundamental frequency of the target child's speech signal within a specific frequency range, and uses the linear predictive coding analysis method to solve for the first three formant frequencies. These frequencies, along with the normalized statistical values, are then appended to the multidimensional voiceprint feature sub-vector to obtain multidimensional voiceprint features, including: The fundamental frequency of the target child's speech signal is extracted within a set frequency range by using the autocorrelation function method combined with center clipping technology. Using the linear predictive coding analysis method, the prediction order is set, and the frequencies of the first three resonant peaks are solved to obtain the resonant peaks; The normalized fundamental frequency mean, variance, and center frequency of the formant are determined and added to the multidimensional voiceprint feature subvector to obtain multidimensional voiceprint features.

7. A child voiceprint adaptive correction system, characterized in that, include: Acquisition unit, used to acquire the target child's speech signal; The feature extraction unit is used to extract multi-dimensional voiceprint features from the target child's speech signal, and perform normalization processing to construct a high-dimensional voiceprint feature vector. The multi-dimensional voiceprint features include MFCC, fundamental frequency, and formants. A matching unit is used to calculate the Euclidean distance between the high-dimensional voiceprint feature vector and the reference high-dimensional voiceprint feature vector, combined with a decision threshold, to make a matching judgment and determine the final child's voiceprint; wherein, the decision threshold is dynamically adjusted using historical matching data to adapt to voice changes, and the reference high-dimensional voiceprint feature vector is a voiceprint template that is progressively updated using a sliding average method with samples that meet the confidence requirements after screening.

8. The child voiceprint adaptive correction system according to claim 7, characterized in that, The matching unit includes: The distance calculation subunit is used to calculate the Euclidean distance between the high-dimensional voiceprint feature vector and the reference high-dimensional voiceprint feature vector. A comparison subunit is used to compare the Euclidean distance with the decision threshold; A sub-unit is defined to determine that the high-dimensional voiceprint feature vector and the reference high-dimensional voiceprint feature vector are successfully matched when the Euclidean distance is not greater than the decision threshold, so as to obtain the final child's voiceprint; and to determine that the high-dimensional voiceprint feature vector and the reference high-dimensional voiceprint feature vector are not successfully matched when the Euclidean distance is greater than the decision threshold.

9. The child voiceprint adaptive correction system according to claim 7, characterized in that, The feature extraction unit includes: The sub-vector extraction sub-unit is used to generate static MFCC features from the target child's speech signal through Mel filter bank, logarithmic energy conversion and discrete cosine transform, and to calculate the first and second order differences to capture the temporal variation characteristics of the speech, forming a multidimensional voiceprint feature sub-vector. The additional subunit is used to extract the fundamental frequency of the target child's speech signal within a specific frequency range using the autocorrelation function method combined with center clipping technology, and to solve the first three formant frequencies using the linear predictive coding analysis method. These frequencies, along with the normalized statistical values, are then added to the multidimensional voiceprint feature subvector to obtain multidimensional voiceprint features. The vector processing subunit is used to normalize the multi-dimensional voiceprint features through cepstral mean variance to eliminate channel effects and individual volume differences, and then define them as the current input and registered template feature vectors to obtain high-dimensional voiceprint feature vectors.

10. The child voiceprint adaptive correction system according to claim 9, characterized in that, The additional subunit includes: The fundamental frequency extraction module is used to extract the fundamental frequency of the target child's speech signal within a set frequency range using the autocorrelation function method combined with center clipping technology. The formant extraction module is used to use the linear predictive coding analysis method, set the prediction order, and solve for the frequencies of the first three formants to obtain the formants. The feature addition module is used to determine the normalized fundamental frequency mean, variance, and center frequency of the formant, and add them to the multidimensional voiceprint feature subvector to obtain multidimensional voiceprint features.

Images