This paper proposes stroke matching based text recognizer (SMTR), a new approach to recognize Bangla handwritten text (BHT) that mainly focuses on off-line Bangla handwritten characters (BHCs) as well as words. For recognition, at first SMTR simplifies any font, size and shape of BHT into a set of strokes. Then it matches these strokes with its own set of strokes of Bangla characters those are already stored in the database. Here the structural properties of lots of these stored strokes are proposed by SMTR. Usually handwritten character (HC) recognition means optical character recognition and herein the accuracy of recognition is not so high because of divergences, variations and characteristics of HCs. Although up to now, lots of recognizers are available only to recognize BHCs, a recognizer that can recognize BHT is almost unavailable. Syntactic method (SM) is one of the most popular methods to recognize BHCs. SMTR proposed in this paper modifies existing strokes of SM and exploits it to recognize BHCs as well as BHT. The experimental results show that SMTR can recognize BHCs and also BHT (i.e. words) even in a noisy environment profoundly.