Michał Kręcichwost

Dynamic time warping in phoneme modeling for fast pronunciation error detection

The presented paper describes a novel approach to the detection of pronunciation errors. It makes use of the modeling of well-pronounced and mispronounced phonemes by means of the Dynamic Time Warping (DTW) algorithm. Four approaches that make use of the DTW phoneme modeling were developed to detect pronunciation errors: Variations of the Word Structure (VoWS), Normalized Phoneme Distances Thresholding (NPDT), Furthest Segment Search (FSS) and Normalized Furthest Segment Search (NFSS). The performance evaluation of each module was carried out using a speech database of correctly and incorrectly pronounced words in the Polish language, with up to 10 patterns of every trained word from a set of 12 words having different phonetic structures. The performance of DTW modeling was compared to Hidden Markov Models (HMM) that were used for the same four approaches (VoWS, NPDT, FSS, NFSS). The average error rate (AER) was the lowest for DTW with NPDT (AER=0.287) and scored better than HMM with FSS (AER=0.473), which was the best result for HMM. The DTW modeling was faster than HMM for all four approaches. This technique can be used for computer-assisted pronunciation training systems that can work with a relatively small training speech corpus (less than 20 patterns per word) to support speech therapy at home.

Quantitative Validation of Gait and Swing Angles Determination from Inertial Signals

The still increasing life length expectancy creates new challenges in the field of senior care. It encourages researchers to provide the nursing homes and senior care assistants with tools that will both, rise an alarm in case of a sudden fall and collect data for long-term diagnosis of the declining motor abilities like the number of steps taken per day or changes in some gait parameters. This paper presents a quantitative validation of a remote system for activity monitoring of the elderly based on inertial sensors. It focuses on features connected to walk quality such as number of steps and the swing angle outlined by an ankle in the sagittal plane during walk. A measurement protocol is proposed, a validation method is described and the obtained results are discussed.

Acoustic Mask for Air Flow Distribution Analysis in Speech Therapy

Quantitative analysis of lateral sigmatism is a challenge for contemporary speech therapy. In the paper a prototype of an acquisition device for voice and lateral air flow registration is described. The proposed acoustic mask enables for registration of signal from 7 microphones placed in different positions in the area around the mouth. Tests considering air flows directed centrally and laterally have been conducted. Artificially generated air streams and lateral speech utterances have been considered. Energy distribution analysis has been performed to distinguish the direction of the interferences. The results confirm that the acoustic mask may be used in diagnosis and speech therapy of lateral sigmatism.

Pronunciation Error Detection Using Dynamic Time Warping Algorithm

In the paper a pronunciation error detection method has been presented, wchich is based on word structural features. A lowcomplexity classifier has been proposed, that is not concentrated on a limited base of error patterns, but is flexible enough to find unspecified mispronunciations. Two classification variants using Dynamic Time Warping (DTW) algorithm has been tested on speech corpus containing recordings of 30 people.

Computer-Aided Evaluation of Sibilants in Preschool Children Sigmatism Diagnosis

In this study an approach to the analysis of sibilant sounds is presented. The method includes the use of the signal spectral measures: amplitudes and frequencies of the fricative formants and the normalized spectral moments. Mel-Frequency Cepstral Coefficients are included in the feature vector as well. The feature sets have been classified using the Support Vector Machine with linear kernel. The experiments have been performed on the speech corpus containing 5 selected words pronounced by 60 preschool children with proper pronunciation or with natural pathologies. The best obtained accuracy of the classification is 94.48 %. The method could be used in a computer-assisted sigmatism diagnosis or therapy system.

CAD of Sigmatism Using Neural Networks

Sigmatism, or lisp, is a common speech pathology defined by the misarticulation of sibilants and commonly appears in preschool-age children. Automated diagnosis from speech data has been used for other disorders, and the use of acoustic features could objectify the diagnosis procedure. 1593 multichannel recordings from 85 young children were subjected to feature extraction and classification using a neural network. The classification performance was evaluated for single and multichannel input as well as multiple feature sets and articulation phases. Multichannel recordings increased the classifier accuracy from 78.75% to 87.27% when using cepstral and spectral features. The introduction of a multichannel acoustic features was shown to increase sigmatism detection accuracy.

Novel Geometric Technique of Ultrasound Probe Calibration

In the study a novel, simple and robust calibration technique of ultrasound probe for image-guided navigation systems has been proposed. The method employs a LEGO phantom with known geometry and is much simpler than other commonly used methods. Additionally, it requires only a single US image. The method was compared to the state-of-the-art method, which is based on a N-wire phantom. Obtained results show an improvement in accuracy and reproducibility between the state-of-the-art and the proposed method.

Development of a Multimodal Image Registration and Fusion Technique for Visualising and Monitoring Chronic Skin Wounds

Chronic skin wounds from diabetes, atherosclerosis, and cancer form a large source of morbidity and medical complications. While individual imaging modalities (e.g. visual images, thermograms, ultrasound) can be useful for monitoring the healing process, their use is limited because of the difficulty acquiring and registering multiple images from different modalities. This paper presents a methodology for image registration using an alignment phantom for grayscale images and thermograms. The registration system achieves a Fiducial Registration Error of 0.61 mm and Mutual Information value of 0.774. Future studies will seek to add additional imaging modalities and improve registration for other areas of the body.

Effect of physical activity on the sequelae of osteoporosis in female residents of residential care facilities

BACKGROUND Osteoporosis is one of the most common diseases that develop with age and cause high morbidity and mortality among elderly people. OBJECTIVES This study was set out to evaluate the influence of a program of modified Sinaki exercises, Nordic walking (NW) and a combination of these physical activities on people with osteoporosis. MATERIAL AND METHODS A sample consisting of 91 women aged 65-98 years living in residential care facilities was randomized into 4 groups. The control group (group 1) received only pharmacological treatment. In the other 3 groups, the same drug therapy was enhanced by a program of modified Sinaki exercises (group 2), Nordic walking (group 3), and Sinaki exercises and Nordic walking applied together (group 4). At baseline and after 12 months of intervention, the participants were assessed for bone density, rib cage mobility, motor abilities, risk of falling (Timed Up and Go Test - TUG, Functional Reach Test - FRT), and locomotor activity (based on pedometer readings). The intervention was completed by 83 participants. RESULTS Bone density (T-score) was higher in all intervention groups and in the control group (p < 0.003). The improvement in rib cage mobility was statistically significant in groups 2 (p < 0.001) and 4 (p < 0.002). Locomotor activity significantly improved in groups 3 (p < 0.000) and 4 (p < 0.000). The post-intervention results of the TUG and FRT tests showed a significantly lower risk of falling in group 4. In groups 1 and 2, the risk was higher, but not statistically significantly, and in group 3, it did not change. CONCLUSIONS Modified Sinaki exercises and Nordic walking significantly improved the mobility of the rib cage, locomotor activity and motor abilities in the women comprising groups 2 and 3, but the best results of the intervention were noted in the group treated with both forms of physical activity.

Prototype measurement system for spatial analysis of speech signal for speech therapy

The paper provides a description of a measuring system for multi-channel speech signal analysis. The aim of the work was to construct a spatial recording station and prepare a processing software.