Abstract— The article presents a method of identifying heavily impaired pronunciations of ‘r’ consonant in Romanian language using time domain signal processing techniques. The study focused on words that contain ‘r’ as the first letter and used signals recorded mainly from children as mispronunciations occur most of the time at young age persons. Keywords— speech processing, impaired speech evaluation. I. INTRODUCTION ERSONS with impaired pronunciation are often ignored when it comes down to developing speech processing based systems. There are numerous speech recognition, speaker recognition, voice coding, and voice control applications useful in many different fields but one can not find many solutions for impaired speech identification. Furthermore speech therapy is complicated and requires many visits to specialized cabinets thus increasing the need of an automated system that can be used at home. More exactly, the speech therapy practice is based on correct diagnosis of the pronunciation problems and on their correction in conformity with a precise methodology. The speech therapist needs a large amount of time to precisely identify not only problematic sounds, but also the problematic phonemic combinations. The correct “emission” of affected sounds is made both isolated and in syllables and words. Because of that, a computer application developed for evaluating mispronunciations can lighten the therapy and simplify the specialist effort, offering an etalon-mean for home exercises. Unfortunately automated methods are sparse and still in the research stage mostly due to the characteristics of the sounds analyzed and also due to the large amount of pronunciations defects existing [1], [2]. Also due to the difference in pronunciations from language to language, research in this field is inevitably divided and results from existing Manuscript received October 9, 2001. This work was supported by CNCSIS project no.846/2009 and by CNCSIS project IDEI_1472/2008 Valentin Velican is with Polytechnic University of Bucharest Department of Applied Electronics and Information Engineering. Ovidiu Grigore is with Polytechnic University of Bucharest Department of Applied Electronics and Information Engineering (corresponding author, phone: +40(0)21.402.4897; fax: +40(0)21.402.4957; e-mail: ovidiu.grigore@ upb.ro). Corina Grigore is with Carol Davila University of Medicine and Pharmacy experiments/systems dedicated to a given language [3], [4] are not necessarily useful in other languages. It is obvious, that a final product capable of correcting all the existing mispronunciations is almost impossible to implement as most likely a specific defect needs a specific identification and correction method. Therefore we concentrated our efforts on developing a method that can be applied in correcting mispronunciations of ‘r’ consonant, in Romanian. We chose to study this defect, also known as rhotacism, as it is one of the most common in Romanian. II. PROBLEM FORMULATION Mispronunciations occur in most of the cases on consonants, sounds that contain less energy than vowels. The main problem encountered is that most of the signal processing techniques developed work best (in the case of speech) on signals packing a good amount of energy – i.e. vowels [5], [6]. Finding common features for the correctly pronounced phoneme becomes rather difficult as the sounds analyzed in our case are consonants. Obviously the incorrectly pronounced phonemes are even harder to evaluate as these, in addition to being altered consonants, categorize themselves in several types (all probably with subtypes defined by different degrees of variation from the standard), decreasing the chances of finding a common feature that could simplify the final evaluation and classification algorithm. III. PROBLEM SOLUTION The ‘r’ phoneme is in Romanian language a hard, rhotic consonant being also one of the most commonly mispronounced sounds. The defect is generally known as “rhotacism” and it can be observed from young age children to adults. In the worst possible utterances the ‘r’ is replaced by other sounds like ‘l’, ‘î’ or can completely miss from words [7]. In the mild, linguistic acceptable mispronunciations, the phoneme is replaced with a guttural ‘r’, resembling the pronunciation of ‘r’ in French. A. Database of Recordings The database used in the study was recorded from 7 children and 5 adults pronouncing words starting with ‘r’ like rac, rană, ramă, rață, etc. – in English: crab, wound, frame, duck. It was taken care to collect sounds that have the same “starting consonant – vowel” combination as transitions from the Pattern Recognition Based Method Used in Identifying Impaired Speech Valentin Velican, Ovidiu Grigore, Corina Grigore P Recent Researches ιn Applied Informatics ISBN: 978-1-61804-034-3 190