Tian Swee Tan

Application of Genetic Algorithm in unit selection for Malay speech synthesis system

Corpus based speech synthesis can produce high quality synthetic speech due to it high sensitivity to unit context. Large speech database is embedded in synthesis system and search algorithm (unit selection) is needed to search for the optimal unit sequence. Speech feature which served as target cost is estimated from the input text. The acoustic parameters which served as join cost are derived from mel frequency cepstral coefficients (MFCCs) and Euclidean distance. In this paper, a new method which is Genetic Algorithm is proposed to search for optimal unit sequence. Genetic Algorithm (GA) is a population based search algorithm that is based on the biological principles of selection, reproduction, crossover and mutation. It is a stochastic search algorithm for solving optimization problem. The speech unit sequence that has minimum join cost will be synthesized into complete waveform data.

Text independent Speaker Identification using Gaussian mixture model

This paper describes text-independent (TI) Speaker Identification (ID) using Gaussian mixture models (GMM). The use of GMM approach is motivated by that the individual Gaussian components of a GMM are shown to represent some general speaker-dependent spectral shapes that are effective for speaker identity modeling. For speaker model training, a fast re-estimation algorithm based on highest likelihood mixture clustering is introduced. In this work, the GMM is evaluated on TI Speaker ID task via series of experiments (model convergence, effect of feature set, number of Gaussian components, and training utterance length on identification rate). The database consisted of Malay clean sentence speech database uttered by 10 speakers (3 female and 7 male). Each speaker provides the same 40 sentences utterances (average length- 3.5s) with different text. The sentences for testing were different from those for training. The GMM achieved 98.4% identification rate using 5 training sentences. The model training based on highest likelihood clustering is shown to perform comparably to conventional expectation-maximization training but consumes much shorter computational time.

Automatic phonetic segmentation of Malay speech database

This paper deals with automatic phonetic segmentation for Malay continuous speech. This study investigates fast and automatic phone segmentation in preparing database for Malay concatenative Text-to-Speech (TTS) systems. A 35 Malay phone set has been chosen, which is suitable for building Malay TTS. The segmentation experiment is based on this phone set. HMM based segmentation approach which uses Viterbi force alignment technique is adapted. We use continuous density HMM (CDHMM) with Gaussian mixture which is performs well in speech recognition to prevent large segmentation errors. Besides, this paper presents an implicit boundary refinement method that is incorporated in the Viterbi phonetic alignment. In this approach, the HMM model is trained with phone tokens with their boundaries extended to the be-side phones. This increases the ability of the HMM in modeling phone boundaries and provides effect of implicit boundary refinement when used in phonetic alignment thus reduce segmentation errors. This approach improves increase the performance of baseline HMM segmentation from 42.39%, 74.83%, 84.34% of automatic boundary marks within error smaller than 5, 15, and 25ms to 47.75%, 76.38%, 85.55%.

Heart murmur diagnostic system (HMDS)

Many heart diseases cause changes in heart sounds and additional murmurs before other signs and symptoms appear. Heart sound auscultation is the primary test conducted by general practitioner (GP) . A heart murmur is an abnormal, extra sound during the heartbeat cycle made by blood flowing through the heart and its valves. Murmurs are characterized by their timing, intensity, pitch, shape and location. Timing refers to whether the murmur occurs during systole, diastole or is continuous throughout the cardiac cycle. This paper describes the diagnosis of heart sounds and heart murmurs using stethoscope based on MFCC-HMM. Diagnosing heart sounds depends on the experience and training of the General Practitioner. Echocardiography is the gold standard alternative for diagnosing heart diseases. Even though it provides a more definitive diagnosis in this respect, it is expensive and not widely available throughout Malaysia especially in local hospitals. For the classification based on the HMM, the continuous cyclic heart sound signal needs to be automatically segmented to obtain isolated cycles of the signal. The ECG signal characteristic of the R to R point is used to determine every one minute cyclesof both ECG and heart sound data. The experiment includes varying the number of states and a number of mixtures. In the classification experiments, the proposed method performed successfully with an accuracy of about 98.9% with 5 states,16 gaussion model.