The recognition of the information area with common format in the non-fixed format questionnaire is the major problem in existing questionnaire layout recognition algorithm. To address those problems, a new approach for questionnaire layout analysis based on regional connectivity and neural networks is proposed. First, a center valid graphics is generated by preprocessing the scanned image firstly. Then, a rapid skew correction algorithm is applied for questionnaire images. Next, many questionnaire rows are obtained by using horizontal projection profile segmentation algorithms. After that, the first connected region for each row is extracted to estimate the existence of form region. Based on the analysis of general questionnaire row and table row, a large amount of possible answers region are generated. Finally, the neural network is used to determine the type of possible information areas. Experiments show that the proposed algorithm can automatically identify common questionnaire.