On the use of first and second derivative approximations for biometric online signature recognition

Marcos Faundez-Zanuy,Moises Diaz

DOI: https://doi.org/10.1007/978-3-031-43085-5_36

2024-06-02

Abstract:This paper investigates the impact of different approximation methods in feature extraction for pattern recognition applications, specifically focused on delta and delta-delta parameters. Using MCYT330 online signature data-base, our experiments show that 11-point approximation outperforms 1-point approximation, resulting in a 1.4% improvement in identification rate, 36.8% reduction in random forgeries and 2.4% reduction in skilled forgeries

Computer Vision and Pattern Recognition

What problem does this paper attempt to address?

The problem that this paper attempts to solve is the optimization of feature extraction methods in online signature biometrics, especially by using first - and second - order derivative approximations (i.e., delta and delta - delta parameters) to improve the recognition rate and reduce the false recognition rate of forged signatures. Specifically, the paper explores the impact of different derivative approximation methods on pattern recognition applications and experimentally verifies the significant improvement of the 11 - point approximation method in terms of recognition rate, random forgery, and skilled forgery.

### Main problems:
1. **Improve the accuracy of online signature recognition**: Improve the performance of the online signature recognition system by improving the feature extraction method, especially by using different derivative approximation methods.
2. **Reduce the false recognition rate of forged signatures**: Especially for random forgeries and skilled forgeries, reduce the false recognition rate by optimizing the feature extraction method.

### Experimental results:
- **Recognition rate**: The recognition rate using the 11 - point approximation method is 1.4% higher than that using the 1 - point approximation method.
- **Random forgery**: The false recognition rate of random forgeries is reduced by 36.8%.
- **Skilled forgery**: The false recognition rate of skilled forgeries is reduced by 2.4%.

### Conclusion:
The paper recommends using a 9 - point or 11 - point window to calculate derivatives because these methods perform best in terms of recognition rate and forgery recognition. When using the 11 - point approximation method, the recognition rate is relatively increased by 1.4%, the false recognition rate of random forgeries is reduced by 36.8%, and the false recognition rate of skilled forgeries is reduced by 2.4%.

### Formula summary:
- **Centroid normalization**:
\[
[\bar{x}, \bar{y}]=\frac{1}{L} \sum_{l = 1}^{L}[x_l, y_l]
\]
\[
x_l=x_l-\bar{x}, \quad y_l=y_l-\bar{y}
\]

- **Delta parameter calculation**:
\[
\text{delta}=\frac{\sum_{k=-M}^{M}k\cdot x[k]}{\sum_{k=-M}^{M}k^2}
\]

- **Simple derivative approximation**:
\[
f_i'=\lim_{h\rightarrow0}\frac{f_i(l + h)-f_i(l)}{h}
\]
For single - sample difference:
\[
f_i'=f_i(l + 1)-f_i(l)
\]

- **Z - score normalization**:
\[
\hat{f}_i=\frac{f_i-\bar{f}_i}{\text{std}(f_i)}
\]

Through these methods, the paper shows how to significantly improve the performance of the online signature biometrics system by optimizing feature extraction.