A signature is a sign in written form, a person's identity for whether a document is correct or not, commonly known as a Biometric system. The Biometric system is the most basic, crucial and considered a superb process for a signature in detecting a person's identification and security. Signature forgery is a fraud that often occurs, causing bigger and longer expenses. For reasons like these, a signature detection system must be able to quickly and accurately recognize genuine and dummy signatures. The purpose of this study was to present the original and dummy signature pattern recognition by grouping the original signature data. In this study, Image Segmentation was used to divide the image into several parts, the K-Means Clustering algorithm to group several parts according to the properties of each object, and Feature Extraction of Texture Patterns and Shape Patterns with Gray Level Co-Occurrence Matrix (GLCM) to obtain feature values such as Entropy, Energy, Homogeneity, Correlation, and Contrast which has resulted in a study to detect genuine and counterfeit signatures. Preliminary results show that the percentage of identification of the signature biometric system developed using Feature Extraction with signature shapes on texture patterns got an average similarity rate of: 92.74%, and signature shapes on shape patterns attained an average similarity rate of: 79.20%. Therefore, the texture extraction pattern can detect the degree of similarity between the original signature and the dummy signature with a higher percentage value compared to the shape extraction pattern. The proposed method can produce better accuracy