In order to be used in a myocardial ischemia monitoring system, an on line QRS complex detector based on Haar wavelet was developed and implemented as a recursive digital filter. The detector performance was determined using the available PhysioNet records of the MIT-BIH arrhythmia and European ST-T databases. The resultant detector is fast in execution, easy to implement, and it does not lead to accumulative error, producing 1.19% and 0.19% error rates with MIT-BIH and European ST-T databases respectively. The morphological distortion caused in averaged beats by the detection errors was estimated using the correlation coefficient and the maximal amplitude difference. Thus, it was concluded that the proposed detector is proper to be used by an ischemia monitoring system and, in general, by any system based on averaged beats.