With the rapid spread of smart phones, the demand for speech recognition systems available in mobile environments nowadays is increasing rapidly. This paper proposes a novel method for Korean affricate recognition which is applicable to a phoneme based Korean speech recognition system. The proposed method is based on the volatility indicator values calculated for each constituting block of the input speech signal. Two methods for grouping the volatility indicator values for recognizing affricates are introduced. One is to detect successive blocks of very high volatilities and the other is to detect rapid variations on volatilities over several blocks. By combining these two methods, the positions at which the affricates are recognized are finally determined. According to the experimental results, it is shown that the proposed method can markedly reduce the error rate that the existing zero crossing rate method generates in terms of both the false reject rate and false accept rate.