The purpose of this study was to investigate diadochokinetic (DDK) characteristics in the subjects with spastic cerebral palsy (CP) by severity. DDK characteristics were measured through rate, regularity, accuracy and consistency in Alternate Motion rate (AMR) and Sequential Motion rate (SMR) tasks. The subjects participated in this study included 27 subjects with spastic CP (mild- 9, moderate- 9, severe- 9) and 9 normal persons who are around 11-20 years old. On the result of this study, rate in AMR was significant difference between all spastic groups and normal group, and rate in SMR was significant difference between normal and mild groups and moderate and severe groups. In regularity of the DDK tasks, severe group had significant difference the other groups. Finally, accuracy and consistency of the DDK tasks exhibited significant difference between all spastic groups and normal group. In conclusion, the subjects with spastic CP have a tendency to produce slow and irregular syllable repetition as severity increases, but to produce inaccurate and inconsistent syllable repetition regardless of severity in the DDK tasks.

The present study investigates the vowels of Bahasa Malaysia and Bahasa Indonesia in terms of the first two formant frequencies and provides a three-dimensional formant chart of vowels by plotting F1, F2, and the frequency of datapoints on 4 different scales: Hz, mel, bark, and the number of ERB. For this study, we recruited 30 male native speakers of Bahasa Malaysia and Bahasa Indonesia (15 each) which include 6 vowels (i, e, a, o, u, e) in various contexts. The three-dimensional formant chart showed geophysics of vowel space, such that mountain peaks stand in particular locations with a higher frequency of occurrence of datapoints. The geophysics of vowel space may shed lights on the perceptual structure of vowel space. The results also showed that vowels in utterance-final positions have a significantly higher F1 and a significantly lower F2 than those in utterance-medial or utterance-initial positions, which means that vowels in utterance-final positions are more back and lower in vowel space than those in utterance-medial or utterance-initial positions.

Maximum likelihood linear regression (MLLR) adaptation experiences severe performance degradation with very tiny amount of adaptation data. Eigenspace- based MLLR, as an alternative to MLLR for fast speaker adaptation, also has a weak point that it cannot deal with the mismatch between training and testing environments. In this paper, we propose a simultaneous fast speaker and environment adaptation based on eigenspace-based MLLR. We also extend the sub-stream based eigenspace-based MLLR to generalize the eigenspace-based MLLR with bias compensation. A vocabulary-independent word recognition experiment shows the proposed algorithm is superior to eigenspace-based MLLR regardless of the amount of adaptation data in diverse noisy environments. Especially, proposed sub-stream eigenspace-based MLLR with bias compensation yields 67% relative improvement with 10 adaptation words in 10 dB SNR environment, in comparison with the conventional eigenspace-based MLLR.

In this paper, a rule-based (transformation-based) NE recognition system is proposed. This system uses Brill's rule inference approach. The performance of the rule-based system and IdentiFinder, one of most successful stochastic systems, are compared. In the baseline case (no punctuation and no capitalisation), both systems show almost equal performance. They also have similar performance in the case of additional information such as punctuation, capitalisation and name lists. The performances of both systems degrade linearly with the number of speech recognition errors, and their rates of degradation are almost equal. These results show that automatic rule inference is a viable alternative to the HMM-based approach to NE recognition, but it retains the advantages of a rule-based approach.

For a large corpus of time-aligned data, HMM based approaches are most widely used for automatic segmentation, providing a consistent and accurate phone labeling scheme. There are two methods for training in HMM. Flat starting method has a property that human interference is minimized but it has low accuracy. Bootstrap method has a high accuracy, but it has a defect that manual segmentation is required In this paper, a new algorithm is proposed to minimize manual work and to improve the performance of automatic segmentation. At first phase, voiced, unvoiced and silence classification is performed for each speech data frame. At second phase, the phoneme sequence is aligned dynamically to the voiced/unvoiced/silence sequence according to the acoustic phonetic rules. Finally, using these segmented speech data as a bootstrap, phoneme model parameters based on HMM are trained. For the performance test, hand labeled ETRI speech DB was used. The experiment results showed that our algorithm achieved 10% improvement of segmentation accuracy within 20 ms tolerable error range. Especially for the unvoiced consonants, it showed 30% improvement.

One key factor that hinders the widespread deployment of speaker identification technologies is the requirement of long enrollment utterances to guarantee low error rate during identification. To gain user acceptance of speaker identification technologies, adaptation algorithms that can enroll speakers with short utterances are highly essential. To this end, this paper applies MLLR speaker adaptation for speaker enrollment and compares its performance against other speaker modeling techniques: GMMs and HMM. Also, to speed up the computational procedure of identification, we apply speaker clustering method which uses principal component analysis (PCA) and weighted Euclidean distance as distance measurement. Experimental results show that MLLR adapted modeling method is most effective for short enrollment utterances and that the GMMs performs better when long utterances are available.

The development of efficient multimodal interfaces and fusion algorithms requires knowledge of usage patterns that show how people use multiple modalities. We analyzed multimodal usage patterns for TV-guide application scenarios (or tasks). In order to collect usage patterns, we implemented a multimodal usage pattern collection system having two input modalities: speech and touch-gesture. Fifty-four subjects participated in our study. Analysis of the collected usage patterns shows a positive correlation between the task type and multimodal usage patterns. In addition, we analyzed the timing between speech-utterances and their corresponding touch-gestures that shows the touch-gesture occurring time interval relative to the duration of speech utterance. We believe that, for developing efficient multimodal fusion algorithms on an application, the multimodal usage pattern analysis for the given application, similar to our work for TV guide application, have to be done in advance.