Biometric identification method, apparatus, storage medium, and device

Abstract

The application discloses a biometric feature recognition method, device, storage medium and equipment, and belongs to the biometric feature recognition field. In the application, when the comparison score reaches the set comparison threshold value and the comprehensive quality score does not reach the set quality threshold value, the level of the set comparison threshold value is improved to obtain an enhanced comparison threshold value, and the level of the set quality threshold value is reduced to obtain a degraded quality threshold value; then the biometric feature recognition is performed again using the enhanced comparison threshold value and the degraded quality threshold value, and when the comparison score reaches the enhanced comparison threshold value and the comprehensive quality score reaches the degraded quality threshold value, it is determined that the biometric feature recognition is passed. The application balances the comparison similarity level and the quality intensity, and improves the pass rate of the biometric feature recognition as much as possible without reducing the reliability.

Description

Technical Field

[0001] This application relates to the field of biometric identification, and in particular to a biometric identification method, apparatus, storage medium and device. Background Technology

[0002] A general biometric identification system mainly consists of two parts: biometric registration and verification (1:1 comparison) / identification (1:N comparison). The registration process includes biometric sample acquisition, sample quality assessment, template creation, and template storage. The verification / identification process includes biometric sample acquisition, sample quality assessment, feature extraction, feature comparison, and decision-making.

[0003] Sample quality refers to the degree to which a biometric sample meets the specified conditions of a target application. It is typically represented by a sample quality score. Sample quality assessment is a fundamental step in the biometric identification process, potentially determining the success or failure of biometric matching. Sample quality assessment is used not only to determine sample eligibility but also to evaluate matching reliability. Generally, better sample quality results in higher recognition strength and greater reliability. A common method for assessing sample quality is to compare the sample quality score with a quality threshold to determine if the score meets the threshold.

[0004] In some biometric identification methods, not only is the comparison score compared with the comparison threshold, but the sample quality score is also compared with the quality threshold. Only when both comparisons pass is the biometric identification considered successful.

[0005] However, this biometric identification method has the potential to result in the incorrect discarding or rejection of biometric samples, leading to a lower success rate in biometric identification. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this application provides a biometric identification method, apparatus, storage medium, and device that maximizes the pass rate of biometric identification without compromising reliability.

[0007] The technical solution provided in this application is as follows:

[0008] In a first aspect, this application provides a biometric identification method, the method comprising:

[0009] Two biometric samples are acquired for comparison, and a comprehensive quality score representing the overall quality of the two biometric samples is calculated.

[0010] Biometric comparison is performed on the two biometric samples involved in the comparison to obtain a comparison score;

[0011] The comparison score is compared with a set comparison threshold, and the overall quality score is compared with a set quality threshold.

[0012] When the comparison score reaches the set comparison threshold and the comprehensive quality score reaches the set quality threshold, the biometric identification is deemed successful.

[0013] When the comparison score reaches the set comparison threshold and the overall quality score does not reach the set quality threshold, the level of the set comparison threshold is increased to obtain the enhanced comparison threshold, and the level of the set quality threshold is decreased to obtain the degraded quality threshold.

[0014] The comparison score is compared with the enhanced comparison threshold, and the overall quality score is compared with the degraded quality threshold. When the comparison score reaches the enhanced comparison threshold and the overall quality score reaches the degraded quality threshold, the biometric identification is deemed successful.

[0015] Furthermore, the set quality threshold and the set comparison threshold are obtained by the following method:

[0016] Construct a sample pair dataset containing biometric sample pairs and calculate the overall quality score for each biometric sample pair;

[0017] The biometric sample pairs include homologous sample pairs and non-homologous sample pairs;

[0018] Multiple sample discard rates are set. For each sample discard rate, biometric sample pairs that meet the proportion of the sample discard rate are discarded from the sample pair dataset according to the comprehensive quality score, so as to obtain the target set consisting of undiscarded sample pairs corresponding to each sample discard rate.

[0019] Biometrics are compared for all biometric sample pairs contained in each target set to obtain a set of comparison scores corresponding to the discard rate of each sample.

[0020] The set of comparison scores includes the comparison scores of homologous sample pairs and the comparison scores of non-homologous sample pairs.

[0021] Set at least one false match rate level, and obtain the matching threshold value and false mismatch rate corresponding to the discard rate of each sample under each false match rate level based on each matching score set;

[0022] Based on the sample discard rate and false mismatch rate at each false match rate level, a fitting curve is obtained for each false match rate level.

[0023] Set the desired false mismatch rate for biometric identification, and obtain the quality threshold value for each false mismatch rate level based on the fitted curve at each false mismatch rate level;

[0024] The quality threshold value and comparison threshold value for the selected mismatch rate level are obtained from the quality threshold value and comparison threshold value for each mismatch rate level, and are used as the set quality threshold and the set comparison threshold.

[0025] Furthermore, obtaining the quality threshold value for each false match rate level based on the fitted curve for each false match rate level includes:

[0026] The sample discard rate corresponding to the desired mismatch rate is obtained based on the fitted curve at each mismatch rate level.

[0027] The overall quality score that makes the sample discard rate closest to the sample discard rate value is used as the quality threshold value.

[0028] Furthermore, the step of obtaining the comparison threshold value and false mismatch rate corresponding to the discard rate of each sample under each false match rate level based on each comparison score set includes:

[0029] The comparison threshold value corresponding to the false match rate level is determined based on the comparison score of the non-homologous sample pairs, and the false mismatch rate corresponding to the comparison threshold value is determined based on the comparison score of the homologous sample pairs.

[0030] or;

[0031] Based on the DET curve, obtain the comparison threshold value and false mismatch rate at the false match rate level.

[0032] Furthermore, the construction of a sample pair dataset containing biometric sample pairs and the calculation of a comprehensive quality score for each biometric sample pair include:

[0033] Construct a biometric sample dataset, wherein the biometric sample dataset includes a series of biometric samples;

[0034] The biometric sample dataset contains biometric samples that are combined in pairs to form biometric sample pairs, thus obtaining the sample pair dataset.

[0035] Calculate the sample quality scores of the two biometric samples contained in the biometric sample pair, and calculate the comprehensive quality score of the biometric sample pair based on the sample quality scores of the two biometric samples contained in the biometric sample pair.

[0036] Furthermore, increasing the level of the set comparison threshold to obtain an enhanced comparison threshold includes:

[0037] Select a fitting curve with a level higher than the selected mismatch rate level as the target curve, and obtain the corresponding comparison threshold value on the target curve according to the sample discard rate value corresponding to the set quality threshold, which is used as the enhanced comparison threshold.

[0038] Furthermore, reducing the level of the set quality threshold to obtain a degraded quality threshold includes:

[0039] Under the false match rate level corresponding to the target curve, biometric sample pairs whose comparison scores in the sample pair dataset reach the comparison threshold value under the false match rate level corresponding to the target curve and whose comprehensive quality scores do not reach the set quality threshold are counted as erroneously discarded sample pairs.

[0040] Statistical analysis is performed on the overall quality scores of all the erroneously discarded sample pairs to obtain the downgraded quality threshold.

[0041] Secondly, this application provides a biometric identification device, the device comprising:

[0042] The quality score calculation module is used to acquire two biometric samples for comparison and calculate a comprehensive quality score that represents the overall quality of the two biometric samples for comparison.

[0043] The comparison score calculation module is used to perform biometric comparison on the two biometric samples participating in the comparison and obtain a comparison score.

[0044] The comparison module is used to compare the comparison score with a set comparison threshold, and to compare the overall quality score with a set quality threshold;

[0045] The first determination module is used to determine that biometric identification is successful when the comparison score reaches the set comparison threshold and the comprehensive quality score reaches the set quality threshold.

[0046] The threshold adjustment module is used to increase the level of the set comparison threshold to obtain an enhanced comparison threshold and decrease the level of the set quality threshold to obtain a degraded quality threshold when the comparison score reaches the set comparison threshold and the overall quality score does not reach the set quality threshold.

[0047] The second determination module is used to compare the comparison score with the enhanced comparison threshold and the comprehensive quality score with the degraded quality threshold. When the comparison score reaches the enhanced comparison threshold and the comprehensive quality score reaches the degraded quality threshold, the biometric identification is determined to be successful.

[0048] Furthermore, the set quality threshold and the set comparison threshold are obtained through the following process:

[0049] The dataset building module is used to build a sample pair dataset containing biometric sample pairs and to calculate the overall quality score for each biometric sample pair.

[0050] The biometric sample pairs include homologous sample pairs and non-homologous sample pairs;

[0051] The discard module is used to set multiple sample discard rates. For each sample discard rate, based on the comprehensive quality score, it discards biometric sample pairs that meet the proportion of the sample discard rate from the sample pair dataset, thereby obtaining a target set consisting of undiscarded sample pairs corresponding to each sample discard rate.

[0052] The comparison module is used to perform biometric comparison on all biometric sample pairs contained in each target set to obtain a set of comparison scores corresponding to the discard rate of each sample.

[0053] The set of comparison scores includes the comparison scores of homologous sample pairs and the comparison scores of non-homologous sample pairs.

[0054] The comparison performance parameter determination module is used to set at least one false match rate level, and obtain the comparison threshold value and false mismatch rate corresponding to the discard rate of each sample under each false match rate level based on each comparison score set;

[0055] The curve fitting module is used to fit a fitting curve for each level of false match rate based on the sample discard rate and false mismatch rate.

[0056] The quality threshold determination module is used to set the desired false mismatch rate for biometric recognition and obtain the quality threshold value for each false mismatch rate level based on the fitted curve at each false mismatch rate level.

[0057] The selection module is used to obtain the quality threshold value and comparison threshold value of the selected mismatch rate level from the quality threshold value and comparison threshold value of each mismatch rate level, and use them as the set quality threshold and the set comparison threshold.

[0058] Furthermore, the quality threshold value for each false match rate level is obtained based on the fitted curve for each false match rate level through the following process:

[0059] The sample discard rate corresponding to the desired mismatch rate is obtained based on the fitted curve at each mismatch rate level.

[0060] The overall quality score that makes the sample discard rate closest to the sample discard rate value is used as the quality threshold value.

[0061] Furthermore, the comparison performance parameter determination module includes:

[0062] The first processing unit is used to determine the comparison threshold value corresponding to the mismatch rate level based on the comparison score of the non-homologous sample pairs, and to determine the mismatch rate corresponding to the comparison threshold value based on the comparison score of the homologous sample pairs.

[0063] Alternatively, the comparison performance parameter determination module includes:

[0064] The second processing unit is used to obtain the comparison threshold value and the false mismatch rate at the false match rate level based on the DET curve.

[0065] Furthermore, the dataset construction module includes:

[0066] A sample dataset construction unit is used to construct a biometric sample dataset, wherein the biometric sample dataset includes a series of biometric samples;

[0067] The sample pair dataset construction unit is used to combine the biofeature samples contained in the biofeature sample dataset in pairs to form biofeature sample pairs, thereby obtaining the sample pair dataset;

[0068] The comprehensive quality score calculation unit is used to calculate the sample quality score of the two biometric samples contained in the biometric sample pair, and to calculate the comprehensive quality score of the biometric sample pair based on the sample quality scores of the two biometric samples contained in the biometric sample pair.

[0069] Furthermore, the threshold adjustment module includes:

[0070] The enhanced alignment threshold determination unit is used to select a fitting curve with a level higher than the selected mismatch rate level as the target curve, and obtain the corresponding alignment threshold value on the target curve according to the sample discard rate value corresponding to the set quality threshold, which is used as the enhanced alignment threshold.

[0071] Furthermore, the threshold adjustment module includes:

[0072] The unit for determining erroneously discarded sample pairs is used to count biometric sample pairs whose comparison scores in the sample pair dataset reach the comparison threshold value at the mismatch rate level corresponding to the target curve and whose overall quality score does not reach the set quality threshold, and these pairs are then identified as erroneously discarded sample pairs.

[0073] The degradation quality threshold determination unit is used to perform statistical analysis on the comprehensive quality scores of all the erroneously discarded sample pairs to obtain the degradation quality threshold.

[0074] Thirdly, this application provides a computer-readable storage medium for biometric identification, including a memory for storing processor-executable instructions, which, when executed by the processor, implement the steps of the biometric identification method described above.

[0075] Fourthly, this application provides a device for biometric identification, including at least one processor and a memory storing computer-executable instructions, wherein the processor executes the instructions to implement the steps of the biometric identification method.

[0076] This application has the following beneficial effects:

[0077] In this application, for cases where "the sample quality strength is insufficient but the comparison similarity level is sufficient", the method of "increasing the comparison threshold and decreasing the quality threshold" is adopted. Then, the enhanced comparison threshold and the degraded quality threshold are used to perform biometric identification. By balancing the comparison similarity level and quality strength, the pass rate of biometric identification is increased as much as possible without reducing reliability. Attached Figure Description

[0078] Figure 1 This is a flowchart illustrating the biometric identification method of this application;

[0079] Figure 2 A schematic diagram of the fitting curves at different FMR levels;

[0080] Figure 3 This is a schematic diagram of the biometric identification device of this application. Detailed Implementation

[0081] To make the technical problems, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with the accompanying drawings and specific embodiments. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0082] This application provides a biometric identification method, such as... Figure 1 As shown, the method includes:

[0083] S100: Obtain two biometric samples for comparison and calculate the overall quality score representing the combined quality of the two biometric samples for comparison.

[0084] Sample quality refers to the degree to which a biometric sample meets the specified conditions of a target application, and is generally represented by a sample quality score. Typically, a higher sample quality score indicates better sample quality. However, in some cases, lower scores may be used to represent higher quality, meaning a smaller sample quality score indicates better sample quality. In one example, the quality score is between 0 and Q. Max (Q) Max Generally, the value is between 100 and 100. In biometric identification, the sample quality score should reach the set threshold Q. t .

[0085] The overall quality score of two biometric samples is calculated by combining the sample quality scores of the two biometric samples. For example, the minimum, average, or other statistical values ​​of the two samples can be used, and this application does not impose any restrictions on this.

[0086] S200: Perform biometric comparison on two biometric samples to obtain a comparison score.

[0087] S300: Compare the comparison score with the set comparison threshold, and compare the overall quality score with the set quality threshold.

[0088] S400: When the comparison score reaches the set comparison threshold and the overall quality score reaches the set quality threshold, the biometric identification is deemed successful.

[0089] In this application, the similarity between two biometric samples is determined not only by comparing their scores with a set comparison threshold, but also by comparing their overall quality score with a set quality threshold to determine the quality of the two biometric samples and make a decision on the strength of biometric recognition.

[0090] Strength decision-making is the process of determining whether biometric identification results are sufficiently acceptable. For example, a high alignment score, or measurement of sample quality, generally indicates better quality and alignment performance, showing a positive correlation: "higher quality scores for homologous samples result in higher alignment similarity scores, while higher quality scores for non-homologous samples result in lower alignment similarity scores." This principle can serve as the theoretical basis for strength decision-making or reliability assessment.

[0091] In view of the above understanding, this application provides a biometric identification method capable of reliability assessment, which mainly evaluates the reliability of biometric comparison results by assessing the quality of the two biometric samples involved in the comparison. After biometric sample comparison, it is necessary not only to determine the comparison score and perform biometric matching to determine the comparison similarity level, but also to perform an intensity decision to determine the sample quality intensity. The intensity decision requires the comprehensive quality score to reach a set quality threshold. From the perspective of sample pair quality intensity, the reliability of the comparison results is evaluated, solving the problem of the singularity of judging the matching result solely based on the comparison score.

[0092] In the aforementioned biometric identification process, if the overall quality score is low and does not reach the set quality threshold, but the comparison score exceeds the set comparison threshold, i.e., "the sample quality strength is insufficient, but the comparison similarity level is sufficient," then the biometric identification is deemed unsuccessful. In this case, the biometric sample participating in the comparison may be incorrectly discarded or rejected, which may actually be a misjudgment.

[0093] To address the above problems, this application also includes the following steps:

[0094] S500: When the comparison score reaches the set comparison threshold and the overall quality score does not reach the set quality threshold, the level of the set comparison threshold is increased to obtain the enhanced comparison threshold, and the level of the set quality threshold is decreased to obtain the degraded quality threshold.

[0095] The enhanced matching threshold mentioned in this application refers to a threshold that makes biometric matching more difficult to pass, while the degraded quality threshold refers to a threshold that makes intensity decisions easier to pass, rather than simply referring to numerical values.

[0096] S600: Compare the comparison score with the enhanced comparison threshold and the comprehensive quality score with the degraded quality threshold. When the comparison score reaches the enhanced comparison threshold and the comprehensive quality score reaches the degraded quality threshold, the biometric recognition is deemed successful.

[0097] In this application, for cases where "sample quality strength is insufficient but comparison similarity level is sufficient" (i.e., the overall quality score does not reach the set quality threshold, but the comparison score reaches the set comparison threshold), the method of "increasing the comparison threshold and decreasing the quality threshold" is adopted. Then, the enhanced comparison threshold and the degraded quality threshold are used to perform biometric identification. By balancing the comparison similarity level and quality strength, the pass rate of biometric identification is increased as much as possible without reducing reliability.

[0098] The aforementioned threshold settings for comparison, quality, enhanced comparison, and degraded quality can be set manually in advance. However, manual setting relies on experience, resulting in low rationality and reliability.

[0099] To address the aforementioned issues, the set quality threshold and set comparison threshold in this application can be obtained using the following method.

[0100] S1: Construct a dataset of sample pairs containing biometric sample pairs and calculate the overall quality score for each biometric sample pair.

[0101] Among them, biometric sample pairs include homologous sample pairs and non-homologous sample pairs.

[0102] A biometric sample pair refers to a combination of two biometric samples used in biometric identification. Biometric sample pairs may be paired (samples from the same source) or unpaired (samples from different sources). The overall quality score is calculated by combining the sample quality scores of the two biometric samples contained in the biometric sample.

[0103] S2: Set multiple sample discard rates. For each sample discard rate, discard biometric sample pairs that meet the proportion of the sample discard rate from the sample pair dataset according to the comprehensive quality score, and obtain the target set consisting of undiscarded sample pairs corresponding to each sample discard rate.

[0104] The sample discard rate (SDR) refers to the ratio of samples whose quality scores do not meet a relevant threshold. In this application, it refers to the ratio of biometric sample pairs whose overall quality scores do not meet the relevant threshold.

[0105]

[0106] In the formula: S1 is the number of biometric sample pairs whose overall quality score does not meet the relevant discard threshold, and S2 is the number of all biometric sample pairs in the sample pair dataset.

[0107] When the overall quality score is higher, indicating better quality of biometric sample pairs, S1 represents the number of biometric sample pairs with an overall quality score below the relevant discard threshold. Conversely, when the overall quality score is lower, indicating better quality of biometric sample pairs, S1 represents the number of biometric sample pairs with an overall quality score above the relevant discard threshold.

[0108] Taking the example that a higher overall quality score indicates better quality biometric sample pairs, when discarding samples, they can be sorted from highest to lowest overall quality score. Then, according to multiple set discard rates (e.g., discard rate = 0.1, 0.2, 0.3…), biometric sample pairs with lower overall quality scores can be discarded proportionally (e.g., 0.1, 0.2, 0.3…). The remaining biometric sample pairs form a target set, with each discard rate (e.g., 0.1, 0.2, 0.3…) corresponding to one target set.

[0109] S3: Perform biometric comparison on all biometric sample pairs contained in each target set to obtain a set of comparison scores corresponding to the discard rate of each sample.

[0110] The alignment score set includes alignment scores for homologous sample pairs and alignment scores for non-homologous sample pairs. The alignment scores for homologous sample pairs and non-homologous sample pairs are used to calculate subsequent FNMR and FMR, respectively.

[0111] In this application, the alignment scores of homologous sample pairs and the alignment scores of non-homologous sample pairs can be saved in list form. In this way, for each sample discard rate (e.g., 0.1, 0.2, 0.3, etc.), a list of the corresponding alignment scores of homologous sample pairs and the alignment scores of non-homologous sample pairs is obtained.

[0112] S4: Set at least one false match rate level, and obtain the comparison threshold value and false mismatch rate corresponding to the discard rate of each sample under each false match rate level based on each comparison score set.

[0113] The performance of biometric recognition algorithms is reflected in metrics such as false match rate (FNMR) and false non-match rate (FNMR). The evaluation of these performance metrics is conducted on the aforementioned dataset of sample pairs. Evaluating the FNMR requires comparison between homologous sample pairs, while evaluating the FMR requires comparison between non-homologous sample pairs.

[0114] The false match rate (FMR) is the percentage of samples from different people (different sources) that are mistakenly identified as belonging to the same person (same source).

[0115]

[0116] in, This represents the total number of non-homologous sample pairs. This represents the number of times non-homologous sample pairs are identified as the same person.

[0117] The false non-match rate (FNMR) is the percentage of samples from the same person (from the same source) that are mistakenly identified as samples from different people (from different sources).

[0118]

[0119] in, This represents the total number of homologous sample pairs. This represents the number of times that a pair of samples from the same source is identified as belonging to different people.

[0120] There is a correlation between FMR, FNMR, and the alignment threshold. Under different alignment thresholds, the values ​​of FMR and FNMR are different, and there is a definite correspondence. Therefore, after setting the FMR level, the corresponding alignment threshold value and FNMR value can be determined based on the alignment score set.

[0121] For each FMR level (e.g., FMR=0.001, 0.0001, etc.), there are multiple sample discard rates (e.g., 0.1, 0.2, 0.3...). Each sample discard rate (e.g., 0.1, 0.2, 0.3...) corresponds to a target set, that is, a set of alignment scores. Each set of alignment scores corresponds to an alignment threshold value and an FNMR value.

[0122] Therefore, for an FMR level, there are multiple SDR, FNMR, and alignment threshold values.

[0123] S5: Based on the sample discard rate and false mismatch rate at each false match rate level, fit the fitting curve at each false match rate level.

[0124] For example, such as Figure 2 As shown, an FNMR-SDR curve can be plotted with SDR as the x-axis and FNMR as the y-axis to obtain the fitted curve. Each mismatch rate level (e.g., FMR=0.001, 0.0001, etc.) corresponds to a fitted curve.

[0125] Figure 2 In the diagram, curve 1 represents the FNMR-SDR curve when FMR = 0.001, and curve 2 represents the FNMR-SDR curve when FMR = 0.0001. From... Figure 2 As can be seen, for the same FNMR, the lower the FMR, the higher the SDR (sample rejection rate), that is, the higher the quality threshold corresponding to SDR.

[0126] S6: Set the desired mismatch rate for biometric identification, and obtain the quality threshold value for each mismatch rate level based on the fitted curve at each mismatch rate level.

[0127] For example, on the FNMR-SDR curves mentioned above, the corresponding sample rejection rate for each fitted curve is found by determining the desired false mismatch rate (e.g., 0.02, see point a; such a rejection rate is low and does not affect user experience). For instance, on the curve with an expected FMR of 0.001, when the FNMR is 0.02, the SDR is 0.25 (see point a); on the curve with an expected FMR of 0.0001, when the FNMR is 0.02, the SDR is 0.4 (see point b). The quality threshold value Q is then inferred from the sample rejection rate values. tUnder the selected FNMR, multiple FMR levels of mass threshold values ​​can be obtained, forming different levels of mass threshold values, Q. t1 Q t2 Q t3 wait.

[0128] S7: Obtain the quality threshold value and comparison threshold value for the selected mismatch rate level from the quality threshold value and comparison threshold value for each mismatch rate level, and use them as the set quality threshold and set comparison threshold.

[0129] In practical applications, the design of biometric identification systems will determine the selected false match rate level based on the security requirements of the scenario (such as unlocking, payment, access control, etc.), thereby determining the corresponding quality threshold value and comparison threshold value, which are used as the set quality threshold and set comparison threshold.

[0130] This application, under a set false match rate level, statistically analyzes the false mismatch rate at different discard rates, establishing the relationship between discarding low-quality biometric sample pairs and biometric recognition performance. Specifically, it establishes a fitting curve between the discard rate and the false mismatch rate. Based on the fitting curve, it obtains the quality threshold value at the set false match rate level, and provides a method for quantifying the quality intensity of biometric sample pairs. The quality threshold value in this application is determined based on parameters of biometric recognition performance and is related to both the false mismatch rate and the false match rate. Compared with manually set quality thresholds, it has better rationality and reliability, thereby improving the rationality and reliability of biometric recognition.

[0131] In practical applications, this application allows for the selection of quality threshold values ​​and comparison threshold values ​​under different FMR levels as the set quality threshold and set comparison threshold for different security scenarios, in order to adapt to different security requirements.

[0132] For example, taking iris recognition as an example, the quality threshold value and comparison threshold value are obtained from an iris test dataset of no less than 10,000 pairs of samples as follows:

[0133] When FMR is 0.001 (one in a thousand), FNMR is 0.01, the alignment threshold is 70, and the quality threshold is 65 (taking the minimum sample quality in the sample pair).

[0134] When FMR is 0.0001 (parts ten thousand), FNMR is 0.01, the alignment threshold is 80, and the quality threshold is 75.

[0135] The biometric samples used for comparison were P1 and P2. After sample quality assessment, the quality score of P1 was 68 and the quality score of P2 was 72. The minimum value of 68 was taken as the overall quality score.

[0136] If the selected false match rate level is 0.001, assuming that after comparing P1 and P2, the comparison score S 12 The score was 81, exceeding the set comparison threshold of 70. Simultaneously, the overall quality score Q... 12 = 68 also reached the set quality threshold of 65. This comparison achieved a matching degree and reliability level of FMR=0.001, therefore, the biometric identification passed.

[0137] Clearly, if the selected false match rate (FMR) level is 0.0001, although this comparison reached the comparison threshold level of FMR=0.0001, it did not reach the corresponding reliability level. If FMR=0.0001 is selected, the biometric identification cannot be considered successful.

[0138] As an improvement to the embodiments of this application, the quality threshold value can be determined by the following method:

[0139] The sample discard rate value corresponding to the desired mismatch rate is obtained from the fitted curve at each mismatch rate level.

[0140] The overall quality score that makes the sample discard rate closest to the aforementioned sample discard rate value is used as the aforementioned quality threshold value.

[0141] In the specific calculation, each comprehensive quality score can be used sequentially as a discard threshold to calculate the actual sample discard rate. Alternatively, several comprehensive quality scores can be found at the boundary between discarded and not discarded samples to calculate the actual sample discard rate. The comprehensive quality score corresponding to the actual sample discard rate that is closest to the aforementioned sample discard rate value is selected as the quality threshold value.

[0142] The quality threshold value determined by the above method is exactly the value that is closest to the sample discard rate value, thus improving the accuracy of the quality threshold.

[0143] In S4 above, when calculating the matching threshold value and false mismatch rate corresponding to the discard rate of each sample at the false match rate level, it can be calculated in the following way:

[0144] Method 1:

[0145] The false match rate level is determined by comparing the scores of non-homologous sample pairs (e.g., in the aforementioned list format), and the false mismatch rate is determined by comparing the scores of homologous sample pairs (e.g., in the aforementioned list format).

[0146] Method 2:

[0147] Based on the DET curve, obtain the comparison threshold value and false mismatch rate at the false match rate level.

[0148] The DET curve is a commonly used curve in biometric identification. The DET curve plots FMR (False Match Rate) on the x-axis and FNMR (False Non-Match Rate) on the y-axis. In this application, various alignment threshold values ​​can be set (e.g., 0-100, with each integer representing a threshold value). For each set threshold value, a DET curve is fitted using the false match rate and false non-match rate as coordinates. After plotting the DET curve, the FNMR value corresponding to each expected FMR level, as well as the corresponding alignment threshold value, can be located using the DET curve.

[0149] As an example, the aforementioned S1 includes:

[0150] S11: Construct a biometric sample dataset, wherein the biometric sample dataset includes a series of biometric samples.

[0151] S12: Combine the biometric samples contained in the biometric sample dataset in pairs to form biometric sample pairs, and obtain the sample pair dataset.

[0152] The aforementioned biometric sample dataset can be divided into a registration set and a probe set. The number of biometric samples in the registration set and the probe set can be the same (or different), and there are corresponding relationships (both homologous and non-homologous samples), such as 10,000 pairs of iris samples and 100,000 pairs of face samples. Sample comparisons are performed between the registration set and the probe set (e.g., cross-traversal comparison) to obtain a sample pair dataset.

[0153] S13: Calculate the sample quality score of the two biometric samples contained in the biometric sample pair, and calculate the comprehensive quality score of the biometric sample pair based on the sample quality scores of the two biometric samples contained in the biometric sample pair.

[0154] The overall quality score is calculated from the sample quality scores of two biometric samples. For example, the minimum, average, or other statistical values ​​of the two samples can be used, and this application does not impose any restrictions on this.

[0155] This application raises the level of the set comparison threshold, and one implementation of the enhanced comparison threshold includes the following steps:

[0156] S41: Select a fitting curve with a level higher than the selected mismatch rate level as the target curve, and obtain the corresponding comparison threshold value on the target curve according to the sample discard rate value corresponding to the set quality threshold, which is used as the enhanced comparison threshold.

[0157] An example of a fitted curve is as follows: Figure 2 As shown, in Figure 2 In the example of a selected false match rate (FMR) of 0.001, curve 1 shows the expected Q... TFind the corresponding SDR (e.g., SDR=0.25).

[0158] The fitted curve with FMR=0.0001 and a safety level higher than FMR=0.001 was selected as the target curve, i.e., curve 2. The comparison threshold S for curve 2 was... T2 The comparison threshold S above curve 1 T1 Therefore, S T2 This can be considered as a candidate. On curve 2, point c corresponding to SDR=0.25 is found. In comparison, under the same SDR, curve 2 has a higher FNMR, indicating that the matching is more difficult. The alignment threshold value corresponding to point c is used as the enhancement alignment threshold.

[0159] One implementation of this application to reduce the level of a set quality threshold to obtain a downgraded quality threshold includes the following steps:

[0160] S42: Under the mismatch rate level corresponding to the target curve, the biometric sample pairs whose comparison scores in the statistical sample pair dataset reach the comparison threshold value under the mismatch rate level corresponding to the target curve, and whose comprehensive quality score does not reach the set quality threshold, are regarded as misdiscarded sample pairs.

[0161] Specifically, with FMR=0.0001, the overall quality score q < Q can be calculated. T And the comparison score S ≥ S T2 List and number of quality scores for (mistakenly discarded sample pairs).

[0162] The quality scores of mistakenly discarded sample pairs are listed below:

[0163] for (i=1…N)

[0164]

[0165] end

[0166] q i Let Q be the overall quality score of the i-th pair of biometric samples, i=1…N, where N is the total number of biometric samples. T For the set quality threshold, s i S is the comparison score of the i-th pair of biometric samples. T2 H(x) is the comparison threshold value under the mismatch rate level corresponding to the target curve; H(x) is the step function, which equals 0 when x < 0 and equals 1 when x ≥ 0.

[0167] Take non-zero values The values ​​are sorted from smallest to largest to obtain the list of quality scores for the erroneously discarded sample pairs mentioned above.

[0168] The number of erroneously discarded sample pairs is as follows:

[0169]

[0170] S43: Perform statistical analysis on the overall quality scores of all mistakenly discarded sample pairs to obtain the aforementioned downgraded quality threshold.

[0171] For example, you can take the average of the quality scores in the quality score list ( The median or minimum value is used as the quality threshold for downgrading after a lowering of the grade; this value is lower than Q. T .

[0172] In one example, taking iris recognition as an example, the relevant quality thresholds and comparison thresholds are obtained from an iris test dataset of no less than 10,000 pairs of samples as follows:

[0173] When FMR is 0.001 (parts per thousand), FNMR is 0.01, the alignment threshold is 70, and the quality threshold is 65.

[0174] When FMR is 0.0001 (parts ten thousand), FNMR is 0.01, the alignment threshold is 80, and the quality threshold is 75.

[0175] The biometric sample pairs involved in the comparison were P1 and P2. After sample quality assessment, P1 had a quality score of 60, and P2 had a quality score of 68. The comparison score S after P1 and P2 comparison was... 12 It is 83.

[0176] When the selected FMR is 0.001 (one in a thousand), the comparison score of 83 exceeds the set comparison threshold of 70. The overall quality score, whether the minimum value of 60 or the average value of 64, does not reach the set quality threshold of 65. Therefore, it is advisable to "increase the comparison threshold and decrease the quality threshold".

[0177] According to the aforementioned calculation method, the enhancement alignment threshold is 80 (i.e., the alignment threshold value when FMR is 0.0001). The average, median, and minimum comprehensive quality scores of the mistakenly discarded sample pairs corresponding to the quality threshold of 65 are set to 56, 58, and 50, respectively. The downgrade quality threshold adopts the average value of 56.

[0178] First, the comparison score of 83 exceeds the enhanced comparison threshold of 80; second, the overall quality score (minimum 60 or average 64) reaches the downgraded quality threshold of 56. Therefore, this comparison meets the requirements of the strength decision and the biometric identification is successful.

[0179] This application also provides a biometric identification device, such as... Figure 3 As shown, the device includes:

[0180] The quality score calculation module 100 is used to acquire two biometric samples for comparison and calculate a comprehensive quality score that represents the overall quality of the two biometric samples for comparison.

[0181] The comparison score calculation module 200 is used to compare the biometric features of two biometric samples participating in the comparison and obtain the comparison score.

[0182] The comparison module 300 is used to compare the comparison score with a set comparison threshold and to compare the overall quality score with a set quality threshold.

[0183] The first judgment module 400 is used to determine that biometric identification is successful when the comparison score reaches a set comparison threshold and the overall quality score reaches a set quality threshold.

[0184] The threshold adjustment module 500 is used to increase the level of the set comparison threshold to obtain an enhanced comparison threshold and decrease the level of the set quality threshold to obtain a downgraded quality threshold when the comparison score reaches the set comparison threshold but the overall quality score does not reach the set quality threshold.

[0185] The second determination module 600 is used to compare the comparison score with the enhanced comparison threshold and the comprehensive quality score with the degraded quality threshold. When the comparison score reaches the enhanced comparison threshold and the comprehensive quality score reaches the degraded quality threshold, the biometric identification is determined to be successful.

[0186] In this application, for cases where "the sample quality strength is insufficient but the comparison similarity level is sufficient", the method of "increasing the comparison threshold and decreasing the quality threshold" is adopted. Then, the enhanced comparison threshold and the degraded quality threshold are used to perform biometric identification. By balancing the comparison similarity level and quality strength, the pass rate of biometric identification is increased as much as possible without reducing reliability.

[0187] In this application, the set quality threshold and set comparison threshold can be obtained through the following process:

[0188] The dataset building module is used to build a sample pair dataset containing biometric sample pairs and calculate the overall quality score for each biometric sample pair.

[0189] Among them, biometric sample pairs include homologous sample pairs and non-homologous sample pairs.

[0190] The discard module is used to set multiple sample discard rates. For each sample discard rate, based on the comprehensive quality score, biometric sample pairs that meet the proportion of the sample discard rate are discarded from the sample pair dataset, resulting in a target set consisting of undiscarded sample pairs corresponding to each sample discard rate.

[0191] The comparison module is used to perform biometric comparison on all biometric sample pairs contained in each target set to obtain a set of comparison scores corresponding to the discard rate of each sample.

[0192] The comparison score set includes the comparison scores of homologous sample pairs and the comparison scores of non-homologous sample pairs.

[0193] The comparison performance parameter determination module is used to set at least one false match rate level, and obtain the comparison threshold value and false mismatch rate corresponding to the discard rate of each sample under each false match rate level based on each comparison score set.

[0194] The curve fitting module is used to fit a fitting curve for each level of false match rate based on the sample discard rate and false mismatch rate.

[0195] The quality threshold determination module is used to set the desired false mismatch rate for biometric recognition, and obtains the quality threshold value for each false mismatch rate level based on the fitted curve at each false mismatch rate level.

[0196] The selection module is used to obtain the quality threshold value and comparison threshold value for the selected mismatch rate level from the quality threshold value and comparison threshold value for each mismatch rate level, and use them as the setting quality threshold and the setting comparison threshold.

[0197] Specifically, the quality threshold value can be determined based on the sample discard rate value through the following process:

[0198] The sample discard rate value corresponding to the desired mismatch rate is obtained based on the fitted curve at each mismatch rate level.

[0199] The overall quality score that makes the sample discard rate closest to the sample discard rate value is used as the quality threshold value.

[0200] As an improvement to the embodiments of this application, the aforementioned comparison performance parameter determination module may include:

[0201] The first processing unit is used to determine the alignment threshold value corresponding to the mismatch rate level based on the alignment score of non-homologous sample pairs, and to determine the mismatch rate corresponding to the alignment threshold value based on the alignment score of homologous sample pairs.

[0202] Alternatively, the aforementioned comparison performance parameter determination module may include:

[0203] The second processing unit is used to obtain the comparison threshold value and the false mismatch rate at the false match rate level based on the DET curve.

[0204] In one example, the dataset building module of this application includes:

[0205] A sample dataset construction unit is used to construct a biometric sample dataset, wherein the biometric sample dataset includes a series of biometric samples.

[0206] The sample pair dataset construction unit is used to combine the biofeature samples contained in the biofeature sample dataset in pairs to form biofeature sample pairs, thereby obtaining the sample pair dataset.

[0207] The comprehensive quality score calculation unit is used to calculate the sample quality scores of the two biometric samples contained in the biometric sample pair, and to calculate the comprehensive quality score of the biometric sample pair based on the sample quality scores of the two biometric samples contained in the biometric sample pair.

[0208] This application does not limit the specific form of the threshold adjustment module; in one example, it may include:

[0209] The enhanced alignment threshold determination unit is used to select a fitting curve with a level higher than the selected mismatch rate level as the target curve, and obtain the corresponding alignment threshold value on the target curve based on the sample discard rate value corresponding to the set quality threshold, which is used as the enhanced alignment threshold.

[0210] Furthermore, the threshold adjustment module may also include:

[0211] The erroneous sample pair determination unit is used to identify biometric sample pairs whose comparison scores in the sample pair dataset reach the comparison threshold value of the mismatch rate level corresponding to the target curve and whose overall quality score does not reach the set quality threshold, under the mismatch rate level corresponding to the target curve.

[0212] The degradation quality threshold determination unit is used to perform statistical analysis on the comprehensive quality scores of all erroneously discarded sample pairs to obtain the degradation quality threshold.

[0213] The apparatus provided in the above embodiments corresponds one-to-one with the embodiments of the aforementioned methods in terms of its implementation principle and the resulting technical effects. For the sake of brevity, any parts of the apparatus not mentioned in the embodiments can be referred to the corresponding content in the embodiments of the aforementioned methods. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the modules and units described in this apparatus can all be referred to the corresponding processes in the aforementioned method embodiments, and will not be repeated here.

[0214] The biometric identification method described in the above embodiments of this application can implement business logic through a computer program and record it on a storage medium. This storage medium can be read and executed by a computer, achieving the effects of the scheme described in the method embodiments of this specification. Therefore, embodiments of this application also provide a computer-readable storage medium for biometric identification, including a memory for storing processor-executable instructions. When these instructions are executed by a processor, they implement the steps of the biometric identification method of the foregoing embodiments.

[0215] The storage medium may include a physical device for storing information, typically digitizing the information and then storing it using electrical, magnetic, or optical methods. The storage medium may include: devices that store information using electrical energy, such as various types of memory, like RAM and ROM; devices that store information using magnetic energy, such as hard disks, floppy disks, magnetic tapes, magnetic core memory, bubble memory, and USB flash drives; and devices that store information using optical methods, such as CDs or DVDs. Of course, there are other readable storage media, such as quantum memories and graphene memories.

[0216] The storage medium described above may also include other implementation methods according to the description of the method embodiments. The implementation principle and technical effects of this embodiment are the same as those of the foregoing method embodiments. For details, please refer to the description of the relevant method embodiments, which will not be repeated here.

[0217] This application also provides a device for biometric identification. The device may be a standalone computer, or it may include an actual operating device that uses one or more of the methods or embodiments described in this specification. The biometric identification device may include at least one processor and a memory storing computer-executable instructions. When the processor executes the instructions, it implements the steps of any one or more of the biometric identification methods described above.

[0218] The device described above may also include other implementation methods according to the method embodiments. The implementation principle and technical effects of this embodiment are the same as those of the foregoing method embodiments. For details, please refer to the description of the relevant method embodiments, which will not be repeated here.

[0219] Finally, it should be noted that the above-described embodiments are merely specific implementations of this application, used to illustrate the technical solutions of this application, and not to limit them. The protection scope of this application is not limited thereto. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments, or make equivalent substitutions for some of the technical features, within the scope of the technology disclosed in this application; and these modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application. All should be covered within the protection scope of this application. Therefore, the protection scope of this application should be determined by the protection scope of the claims.

Claims

1. A method of biometric identification, characterized in that, The method comprises: obtaining two biological feature samples participating in comparison, and calculating a comprehensive quality score representing comprehensive quality of the two biological feature samples participating in comparison; performing biological feature comparison on the two biological feature samples participating in comparison to obtain a comparison score; comparing the comparison score with a set comparison threshold, and comparing the comprehensive quality score with a set quality threshold; when the comparison score reaches the set comparison threshold and the comprehensive quality score reaches the set quality threshold, determining that biological feature recognition is passed; when the comparison score reaches the set comparison threshold and the comprehensive quality score does not reach the set quality threshold, increasing a level of the set comparison threshold to obtain an enhanced comparison threshold, and reducing a level of the set quality threshold to obtain a degraded quality threshold; comparing the comparison score with the enhanced comparison threshold, and comparing the comprehensive quality score with the degraded quality threshold, when the comparison score reaches the enhanced comparison threshold and the comprehensive quality score reaches the degraded quality threshold, determining that biological feature recognition is passed.

2. The method of claim 1, wherein, The set quality threshold and the set comparison threshold are obtained by the following method: constructing a sample pair data set comprising biological feature sample pairs, and calculating a comprehensive quality score of each biological feature sample pair; wherein the biological feature sample pairs comprise homologous sample pairs and non-homologous sample pairs; setting a plurality of sample discard rates, for each sample discard rate, discarding biological feature sample pairs in the sample pair data set according to the comprehensive quality score in a proportion conforming to the sample discard rate, to obtain a target set corresponding to each sample discard rate and composed of non-discarded sample pairs; performing biological feature comparison on all biological feature sample pairs contained in each target set to obtain a comparison score set corresponding to each sample discard rate; wherein the comparison score set comprises comparison scores of homologous sample pairs and comparison scores of non-homologous sample pairs; setting at least one false match rate level, and obtaining, according to each comparison score set, a comparison threshold value and a false non-match rate corresponding to each sample discard rate at each false match rate level; fitting to obtain a fitting curve at each false match rate level according to the sample discard rate and the false non-match rate at each false match rate level; setting a false non-match rate to be reached by biological feature recognition, and obtaining a quality threshold value at each false match rate level according to the fitting curve at each false match rate level; obtaining, from the quality threshold value and the comparison threshold value at each false match rate level, a quality threshold value and a comparison threshold value at a selected false match rate level as the set quality threshold and the set comparison threshold.

3. The method of claim 2, wherein, The obtaining of the quality threshold value at each false match rate level according to the fitting curve at each false match rate level comprises: obtaining a sample discard rate value corresponding to the false non-match rate to be reached according to the fitting curve at each false match rate level; so that the comprehensive quality score closest to the sample discard rate value is taken as the quality threshold value.

4. The method of claim 2, wherein, The method comprises the following steps: According to the alignment score of the non-homologous sample pair, the alignment threshold value corresponding to the false matching rate level is determined, and according to the alignment score of the homologous sample pair, the false non-matching rate corresponding to the alignment threshold value is determined. Or; According to the DET curve, the alignment threshold value and the false non-matching rate under the false matching rate level are obtained.

5. The method of claim 2, wherein, The method comprises the following steps: Constructing a sample pair data set containing biological feature sample pairs and calculating the comprehensive quality score of each biological feature sample pair comprises the following steps: Constructing a biological feature sample data set, wherein the biological feature sample data set comprises a series of biological feature samples; Combining the biological feature samples contained in the biological feature sample data set in pairs to form biological feature sample pairs, thereby obtaining the sample pair data set; 6. The method according to any of claims 2-5, wherein, Calculating the sample quality score of the two biological feature samples contained in the biological feature sample pair, and calculating the comprehensive quality score of the biological feature sample pair according to the sample quality score of the two biological feature samples contained in the biological feature sample pair. The method comprises the following steps:

7. The method of claim 6, wherein, Selecting a fitting curve with a level higher than the selected false matching rate level as a target curve, obtaining a corresponding alignment threshold value on the target curve according to the sample discard rate value corresponding to the set quality threshold, and taking the alignment threshold value as the enhanced alignment threshold value. The method comprises the following steps: In the false matching rate level corresponding to the target curve, the biological feature sample pairs in the sample pair data set whose alignment scores reach the alignment threshold value under the false matching rate level corresponding to the target curve and whose comprehensive quality scores do not reach the set quality threshold are counted as false discarded sample pairs; 8. A biometric identification device, characterized by Statistically analyzing the comprehensive quality scores of all the false discarded sample pairs to obtain the degraded quality threshold value. The device comprises: A quality score calculation module for obtaining two biological feature samples participating in alignment and calculating a comprehensive quality score representing the comprehensive quality of the two biological feature samples participating in alignment; An alignment score calculation module for performing biological feature alignment on the two biological feature samples participating in alignment to obtain an alignment score; A comparison module for comparing the alignment score with a set alignment threshold value and comparing the comprehensive quality score with a set quality threshold value; A first determination module for determining that biological feature recognition is passed when the alignment score reaches the set alignment threshold value and the comprehensive quality score reaches the set quality threshold value; A threshold adjustment module for increasing the level of the set alignment threshold value to obtain an enhanced alignment threshold value and decreasing the level of the set quality threshold value to obtain a degraded quality threshold value when the alignment score reaches the set alignment threshold value and the comprehensive quality score does not reach the set quality threshold value. a second determining module configured to compare the comparison score with the enhanced comparison threshold and compare the overall quality score with the degraded quality threshold, and determine that the biometric recognition passes when the comparison score reaches the enhanced comparison threshold and the overall quality score reaches the degraded quality threshold.

9. A computer readable storage medium for biometric recognition, characterized in that, A computer program product comprising a memory for storing processor-executable instructions, which, when executed by the processor, implement the steps of the biometric recognition method of any one of claims 1-7.

10. An apparatus for biometric identification, characterized by A computer program product comprising at least one processor and a memory storing computer-executable instructions, which, when executed by the processor, implement the steps of the biometric recognition method of any one of claims 1-7.

Images