Flexible strain sensors have excellent prospects in the fields of medical health monitoring and intelligent human-machine interaction. Considerable research effort is dedicated to continuously optimizing their sensitivity performance to meet application requirements. However, the ability to distinguish and decouple different forms of strain is crucial in determining the accuracy and effectiveness of information extraction from complex strain patterns, which is also one of the challenges presently restricting the practical application of strain sensors. Here, a simple strategy for constructing a paper-based bending strain sensor is introduced, achieved by a commercially available inkjet printer and original ink cartridges to print on coated paper. The strain sensor exhibits high resolution in measuring strain degrees and decoupling strain in different directions. More importantly, a scheme for measuring the horizontal angle of the strain axis is proposed for the first time, which is achieved through a strain sensor array consisting of three sensing units distributed at different horizontal angles. This further promotes the accurate recognition of strain information in multidimensional strain patterns. Combining machine learning algorithms, the strain sensor array demonstrates exceptional accuracy in distinguishing various human upper limb activities. This work provides the necessary research value to advance flexible strain sensors towards practical applications.