Sigmatism is a speech disorder manifesting through incorrect pronunciation of sibilants. In this paper, we present preliminary results from a computer-assisted speech diagnosis method based on a support vector machine to classify the productions of Polish affricate sibilant according to their place of articulation as alveolar/retroflex (correct) or dental (disordered). Speech samples used in the experiments come from 151 Polish preschool children. Features used in the classification task include mel-frequency cepstral coefficients and frication noise features. The experiments aimed to determine if the combination of MFCC features and frication noise features improves the classification accuracy of an SVM rather than using these features separately. Results indicate that incorporating MFCC and frication noise features of analysed acoustic signals into the classifier can improve accuracy in distinguishing normal and dental articulation of affricates in children’s speech.