Haralick feature selection for material rigidity recognition using ultrasound echo

Object classification based on its rigidity requires principally the recognition of its material consistency. Generally, material consistency can be divided into two families, hard material and soft one. In this context, a new approach based on ultrasonic signal for consistency recognition of object materials is proposed. This approach allows distinguishing between the hard and the soft objects. Material consistency determination is based on Haralick's texture features. Then, a feature selection step is considered to select the most discriminative features. Only three Haralick features were used to assess the efficiency classifications models. As there is no affording dataset of ultrasonic signals acquired for material rigidity recognition, we develop our dataset using two ultrasonic sensors. In this context, no previous work has considered such a challenge. The analysis results show that three parameters (entropy, sum of entropy and variance) were found to be effective to discriminate between the two classes of material rigidity. The obtained results show the efficiency of the proposed method.