Information on the gender of a person plays a vital role in crime investigation, authentication and statistical report on the visitors. In this work, fingerprint ridge count and fingertip size are used as the parameters for automatic gender classification. As a novel method, the optimal score assignment (OSA) method is proposed to classify gender. An optimal score is calculated for male and female from the internally collected fingerprint database. Fingerprints are collected under four age groups and all the fingers are scanned. For the fingerprint image ‘I’ for which gender is to be identified, scores are assigned for ridge count and fingertip assuming that the given image is male. A similar calculation is made assuming that the given image is female. Comparing both values, gender is declared. The maximum success rate attained is 88.41% for the age group 18–25 years and a good success rate of 90.11% is achieved for the right hand ring finger. Performance evaluation is made with the earlier findings of the author and other methods.
Ridge patterns exhibit many properties that reflect the biology of individuals. Ridge parameters such as fingerprint ridge count, ridge density, ridge thickness to valley thickness ratio, ridge width and fingerprint pattern types are used for gender determination. Variations in ridge parameters for male and female are found statistically [1, 2]. Also, it is found that dermatoglyphic features differ statistically between the sexes, ethnic groups and age categories. It is proved by various researchers that a fingerprint can be processed for sex determination [1, 3, 4].
The fingerprint samples were collected from the subjects residing in various parts of Tamil Nadu, India. The Fingkey Hamster II scanner is used for sample collection. The fingerprint image is of 8-bit gray level with a size of 300 × 260 and resolution of 500 dpi. An internal database consisting of fingerprints of 403 males and 410 females is used to test the method. All 10 fingers of each subject were scanned and thus in total, 8130 fingerprints were used. The fingerprints were categorized into four age groups, viz., 8–12, 13–18, 18–25 and above 25. For reference purpose, fingers are numbered 1–10 starting from left little finger to right little finger (left little finger 1, left thumb 5, right thumb 6 and right little finger 10).
In this manuscript, automatic gender identification from the fingerprint ridge count (RC) and fingertip size (FTS) using the OSA method is proposed. Initially, core and delta (singular points) are identified. With respect to core and delta, RCs are determined (traditional method) and in addition, ridge counts measured diagonally (at 45° and 135°) with respect to the core points are averaged. Fingertip size is measured as another parameter to find gender. A high possibility of particular values of ridge count and fingertip size for male and female is identified and given a high score. Proportionate scores are assigned to the remaining values of the ridge count and fingertip considering the most occurring ridge count and fingertip as reference. For an unknown fingerprint, different scores are assigned for RC and FTS for male and female. The sum of these two scores is calculated for male and female. If the male score (MS) is higher than the female score (FS), the decision is declared as male, otherwise it is declared as female. The proposed method of gender classification is demonstrated in Fig. 1.
This method of gender identification will be helpful in short listing the suspects and victims from crime scenes and improves the performance of a system which is used for person recognition and human computer interfaces.
This manuscript is organized as follows: the second section briefs the literature of various gender recognition algorithms using a fingerprint. The third section details the OSA method. In this section, singular points detection, ridge count and fingertip size measurements are elaborated. Score assignment procedure is explained and optimal score is assigned to each ridge count and fingertip size. The experimental results and performance analysis are demonstrated in the section “Experimental results”. The section “Conclusion” concludes the proposed work and briefs the future work.
Although the fingerprint plays an essential role in the identification and verification, only a few machine vision methods have been proposed for gender identification. In this section, we have summarized the prior researches in gender classification.
It is demonstrated that the males have a higher ridge breadth than females . Using Bayes’ theorem  on the rolled fingerprint images belonging to the South Indian population, it is found that the fingerprint possessing ridge density < 13 ridges/25 mm2 is most likely to be of male and ridge count > 14 ridges/25 mm2 are most likely to be of female. Using the ridge thickness to valley thickness ratio (RTVTR) and white lines count features , gender was classified. According to them, the female’s fingerprint is characterized by a high RTVTR, while the male’s fingerprint is characterized by low RTVTR. A proposal for the interactive software system  that relives the tedium of visual inspection and standardizes the fingerprint ridge counting procedure is also published. In terms of age, the quality scores of 18–25 age group are good  compared to < 18 and > 25 age groups.
Ridge distance measurement is vital for robust performance of an automated fingerprint identification system (AFIS) irrespective of quality of the images . Also, the traditional spectral analysis method was realized and a novel statistical method was presented for ridge distance estimation . Ridge density in a particular space was used to classify gender using fingerprint and further demonstrated that the females have a higher ridge density compared with males. Geometric and spectral methods were used to estimate fingerprint ridge distance . These methods calculate ridge direction directly. Mathematical characterization of the local frequency of sinusoidal signals and two-dimensional model was proposed  to approximate the ridgeline patterns for ridgeline density estimation in digital images.
Frequency domain analysis of fingerprint  for the identification of gender produces a good classification rate. Gender classification using fingerprints through univariate decision tree  was proposed and a classification rate of 96.28% was achieved. The back-propagation neural network classifier was used to classify the gender  and the classification rate achieved was 92.67%.
This paper demonstrates the identification of gender using the spatial parameters of the fingerprint. In this work, fingerprint ridge count and fingertip size are used as the parameters for automatic gender classification. As a novel method, the OSA method is proposed to classify gender. Information on the gender of a person plays a vital role in crime investigation, authentication and statistical report on the visitors. This method of gender identification will be helpful in short listing the suspects and victims from crime scenes and improves the performance of a system which is used for person recognition and human computer interfaces.
Optimal score assignment (OSA)
Singular points identification
Basically, the fingerprints are categorized as (a) tented arch, (b) left loop, (c) right loop, (d) whorl, (e) plain arch, (f) central pocket, (g) twin loops and (h) accidents. Except the plain arch , each type has one or more core and delta points referred to as singular points. For the plain arch, for calculation purpose, a point in a ridge which has a high peak is chosen as core and a point in the bottom-most ridge, which is almost straight, is chosen as a delta point. The types of fingerprints and its singular points are illustrated in Fig. 2.
The singular point area is defined as a region where the ridge curvature is higher than normal and where the direction of the ridge changes rapidly . These singular points are useful for fingerprint indexing, i.e., for classification of fingerprint types , fingerprint alignment and orientation field modeling [17, 18] and identification or verification. A core point is the turning point of an innermost ridge. In biometrics and fingerprint scanning, core point refers to the central area of a fingerprint. A delta point is a place where a ridge is bifurcated (or) a delta point is a place where two ridges run side by side and diverge .
The ridge count is calculated by counting the number of ridges intervening between the delta and core . In the proposed method, instead of considering counting only between the core and delta, an effort is taken to count the ridges of the entire fingertip.
To enable this, an imaginary line is drawn between core and delta at 135° (referred to as the principal diagonal) and 45° (referred to as the other diagonal) as shown in Fig. 3.
Let ‘a’ be the ridge count between core to delta, ‘b’ be the ridge count in the principal diagonal and ‘c’ be the ridge count in other diagonal. The total ridge count is calculated by Eq. (1).
Ridge counts were determined for all 8130 fingerprints of 403 male and 410 female fingerprints and analyzed. Details of finger-wise ridge count for male and female and the most common count are presented in Table 1.
From Table 1, it is identified that the minimum as well as the maximum RC are greater for male than female. In addition, the most common RC differs between male and female. The average RC values of male and female (for all fingerprints of the database) are compared in the line chart shown in Fig. 4.
Optimal RC score calculation
Ridge counts of all the internal database fingerprints are calculated using MATLAB. Measured fingerprint RCs are listed in an ascending order and the number of each RC is counted. The percentage of occurrence of a particular RC among total fingerprints is determined by Eq. (2) and presented in Table 2.
From Table 2, it is clarified that the ridge count of 31 was found to be 287 times among the 4100 female fingerprints. This is the highest occurrence in comparison with other RC counts. Its occurrence percentage is calculated as 7 and referred to as the maximum occurrence percentage. A maximum score of 10 is assigned for this RC. Scores for the remaining RC are determined by Eq. (3).
Thus, RC scores are computed individually for male (4030 samples) and female (4100 samples) of all the internal databases and shown in Table 2.
From Table 2, it is concluded that, for female, the RC of 31 is occurring more and, for male, RC of 34 is occurring more. Here, a maximum score of 10 is assigned for each RC.
Fingertip size of the fingerprint
FTS is computed using the scanner information. The scanned image is of the size 300 × 260. In all the 8130 fingerprints of 403 males and 410 females, FTS are figured and analyzed. Comparison of FTS values is made between genders and all four age groups. The fingertip size is computed in square millimeter. The fingertip size of male and female fingerprints irrespective of the age group is analyzed and represented in Table 3.
The FTS values of male and female are compared in the line chart in Fig. 5.
Optimal FTS score calculation
The fingertip size of all the internal database fingerprints is calculated using MATLAB. The measured fingerprint FTSs are listed in ascending order and the number of each FTS is counted. The percentage occurrence of a particular FTS among total fingerprints is determined by Eq. (4) and tabulated in Table 4.
From Table 4, it is clarified that the fingertip size of 450 mm2 was found 175 times among the 4100 female fingerprints. This is the highest occurrence in comparison with other FTS counts. Its occurrence percentage is calculated as 4.2683 and referred to as the maximum occurrence percentage. A maximum score of 10 is assigned for this FTS. Scores for the remaining FTS are determined by Eq. (5).
For example, as in Table 4, FTS = 400 has its occurrence percentage as 2.8780. Now, the score is calculated as follows.
Thus, FTS scores are computed individually for male (4030 samples) and female (4100 samples) of all the internal databases and shown in Table 4.
From Table 4, it is concluded that the FTS of 450 mm2 and 530 mm2 are occurring more for female and male, respectively.
A detailed analysis of RC and FTS was carried out in the previous section. From the analysis, it is observed that the ridge count and the fingertip size of the fingerprints are more for male than female. Also, all these values differ for male and female in all the age groups. The novel method of OSA is discussed and the scores are assigned in this section. As the scores assigned for a particular value of RC/FTS are different for male and female, the sum of these scores computed for each gender is distinguishable and thus declares more accurate results.
Let I be the fingerprint image for which the gender needs to be identified. Considering I as the male fingerprint, the total score IMS is calculated by Eq. (6).
where RCM and FTSM are the respective scores of ridge count and fingertip size assigned for male fingerprints. Similarly, considering I as a female fingerprint, the total score of IFS is calculated by Eq. (7).
where RCF and FTSF are the respective scores of ridge count and fingertip size assigned for female fingerprints. The gender of the unknown fingerprint I is declared as male if IMS > IFS, and otherwise declared as female. Two examples are shown in the Table 5.
Age group-wise gender classification
Age group-wise gender classifications are presented in Table 5. The number of samples used is 44, 55, 198 and 106 for each finger in the age groups 8–12, 13–18, 19–25 and > 25, respectively. Thus, collectively 4030 samples were used for testing the proposed method. For the age group 19–25 years, the results are good and the success rate achieved is 88.41%. The success rate for the right hand ring fingers in this group achieved is 90.11%.
In this section, a novel approach of the gender classification using the OSA method is compared with various methods experimented by the author [12, 20, 21]. The best results were obtained for the age group 19–25 years alone compared with the earlier publications of the author in Table 6.
From the results shown in Table 7, it is observed that the OSA method results (age group 19–25 years) are good individually for male and female. As an overall result, the classification rate achieved is 88.41%. Figure 6 illustrates the increase in classification rate (%) from the frequency domain technique to the spatial parameters technique.
Badawi et al.  compared RTVTR, ridge count, white lines count, ridge count asymmetry and pattern type concordance as features. FCM, LDA, and NN classifiers were used for gender classification. For this study, the RTVTR, and white lines count features were analyzed for 255 persons (150 males, and 105 females). Table 7 shows only the overall classification rate obtained by Badawi et al. , and the proposed method (age group of 19–25). Verma and Agarwal  used ridge density and ridge width in addition to RTVTR as features and with the SVM classifier, the results obtained by them are shown in Table 7. They used a dataset of 400 fingerprints (200 males and 200 females) of Indian origin in the age group of 18–60 years. These fingerprints were divided equally for training and testing with SVM classifier.
A novel method of OSA technique was proposed for gender classification using the ridge count and fingertip size. Performance evaluation was done with the methods tested and the earlier methods by other researchers.
For the proposed method, the spatial parameters, ridge count and fingertip size, and the OSA method were used for gender classification. An extensive analysis of both parameters was done and it is found that all the values obtained are greater for male than female. An algorithm for assigning score for each value of the parameters was discussed. This method produced a success rate of 88.41% and 90.11% is achieved for the right hand ring finger. A comparative performance evaluation was carried out with the other methods tested by the present researchers. Thus, the proposed method achieves better results than all the methods discussed. Also, the OSA method works well even for the poor quality fingerprints. To improve the success rate further, other fingerprint features can also be included.
Kralík M, Novotny V (2003) Epidermal ridge breadth: an indicator of age and sex in paleodermatoglyphics. Var Evol 11:5–30
Cummins H (1941) Ancient finger prints in clay. Sci Mon 52(5):389–402
Nithin MD, Manjunatha B, Preethi DS, Balaraj BM (2011) Gender differentiation by finger ridge count among south Indian population. J For Leg Med 18(2):79–81
Badawi A, Mahfouz M, Tadross R, Jantz, R (2006) Fingerprint-based gender classification. In: The international conference on image processing. Computer vision and pattern recognition 1: 41–46
Abdullah SF, Abas ZA, Rahman AFNA, Saad WHM (2016) Development of a fingerprint algorithm using fingerprint global features. Int J Adv Comput Sci Appl 7(6):275–279
Modi KS, Elliott SJ, Whetsone J, Kim H (2007) Impact of age groups on fingerprint recognition performance. IEEE Workshop Autom Identif Adv Technol 1:19–23
Yin Y, Tian J, Yang X (2004) Ridge distance estimation in fingerprint images: algorithm and performance evaluation. EURSIP J Appl Sig Process 4:495–502
Acree AM (1999) Is there a gender difference in fingerprint ridge density. For Sci Int 102(1):35–44
Hong L, Wan Y, Jain A (1998) Fingerprint image enhancement: algorithm and performance evaluation. IEEE Trans Pattern Anal Mach Intell 20(8):777–789
Kovacs-Vajna MZ, Rovatti R, Frazzoni M (2000) Fingerprint ridge distance computation methodologies. Pattern Recognit 33(1):69–80
Maltoni D, Maio D, Jain AK, Prabhakar S (2003) Handbook of fingerprint recognition, 1st edn. Springer, New York
Gnanasivam P, Muttan S (2011) Gender identification using fingerprint through frequency domain analysis. Eur J Sci Res 59(2):191–199
Samta Gupta, Prabhakar Rao A (2014) Fingerprint based gender classification using discrete wavelet transform & artificial neural network. Int J Comput Sci Mob Comput 3(4):1289–1296
Li J, Yau W-Y, Wang H (2008) Combining singular points and orientation Image information for fingerprint classification. Pattern Recognit 41(1):353–366
Boer JD, Bazen AM, Gerez SH (2001) Indexing fingerprint database based on multiple features. In: Proceedings of ProRISC 2001 workshop on circuits, systems and signal processing, pp 300–306
Zhang Q, Yan H (2004) Fingerprint classification based on extraction and analysis of singularities and pseudo ridges. Pattern Recognit 37(11):2233–2243
Gu J, Zhou J (2003) A novel model for orientation field of fingerprints. Proc IEEE Comput Soc Conf Comput Vis Pattern Recognit 2:493–498
Sherlock B, Monro D (1993) A model for interpreting fingerprint topology. Pattern Recognit 26(7):1047–1055
Hoover JE (2006) The Science of fingerprints classification and uses. Federal Bureau of Investigation. United States Department of Justice. http://www.dermatoglyphics.com/sof/. Accessed Nov 2018
Gnanasivam P, Muttan S (2012) Fingerprint gender classification using wavelet transform and singular value decomposition. Int J Comput Sci Issues 9(3):274–282
Gnanasivam P, Muttan S (2012) Singular value decomposition based fingerprint gender classification. CiiT Int J Biomet Bioinform 4(1):23–27
Verma M, Agarwal S (2009) Fingerprint based male-female classification. CISIS 2008. ASC 53:251–257
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
About this article
Cite this article
Gnanasivam, P., Vijayarajan, R. Gender classification from fingerprint ridge count and fingertip size using optimal score assignment. Complex Intell. Syst. 5, 343–352 (2019). https://doi.org/10.1007/s40747-019-0099-y
- Gender classification
- Optimal score assignment
- Ridge count
- Fingertip size