• Journal of KIISE:Software and Applications
• /
• v.29 no.3
• /
• pp.167-179
• /
• 2002

In this paper, we propose a new acoustic model for characterizing segmental features and an algorithm based upon a general framework of hidden Markov models (HMMs) in order to compensate the weakness of HMM assumptions. The segmental features are represented as a trajectory of observed vector sequences by a polynomial regression function because the single frame feature cannot represent the temporal dynamics of speech signals effectively. To apply the segmental features to pattern classification, we adopted segmental HMM(SHMM) which is known as the effective method to represent the trend of speech signals. SHMM separates observation probability of the given state into extra- and intra-segmental variations that show the long-term and short-term variabilities, respectively. To consider the segmental characteristics in acoustic model, we present segmental-feature HMM(SFHMM) by modifying the SHMM. The SFHMM therefore represents the external- and internal-variation as the observation probability of the trajectory in a given state and trajectory estimation error for the given segment, respectively. We conducted several experiments on the TIMIT database to establish the effectiveness of the proposed method and the characteristics of the segmental features. From the experimental results, we conclude that the proposed method is valuable, if its number of parameters is greater than that of conventional HMM, in the flexible and informative feature representation and the performance improvement.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.38 no.5
• /
• pp.567-575
• /
• 2001

This paper suggests new method to get fixed size parameter from different length of voice signals. The efficiency of speech recognizer is determined by how to compare the similarity(distance of each pattern) of the parameter from voice signal. But the variation of voice signal and the difference of speech speed make it difficult to extract the fixed size parameter from the voice signal. The method suggested in this paper is to normalize the parameter at fixed size by using the 2 dimension DCT(Discrete Cosine Transform) after representing the parameter by spectrogram. To prove validity of the suggested method, parameter extracted from 32 auditory filter-bank(it estimates auditory nerve firing probabilities) is used for the input of neural network after being processed by 2 dimension DCT. And to compare with conventional methods, we used one of conventional methods which solve time alignment problem. The result shows more efficient performance and faster recognition speed in the speaker dependent and independent isolated word recognition than conventional method.

• The KIPS Transactions:PartB
• /
• v.9B no.2
• /
• pp.215-222
• /
• 2002

This paper presents a reliable fully automatic labeling system which fits well with languages having well-developed syllables such as in Korean. The ASL System utilize DAC (Divide and Conquer), a control mechanism, based segmentation algorithm to use phonetic and acoustic information with greater efficiency. The segmentation algorithm is to devide speech signals into speechlets which is localized speech signal pieces and to segment each speechlet for speech boundaries. While HMM method has uniform and definite efficiencies, the suggested method gives framework to steadily develope and improve specified acoustic knowledges as a component. Without using statistical method such as HMM, this new method use only phonetic-acoustic information. Therefore, this method has high speed performance, is consistent extending the specific acoustic knowledge component, and can be applied in efficient way. we show experiment result to verify suggested method at the end.

• The Journal of the Acoustical Society of Korea
• /
• v.22 no.2
• /
• pp.135-144
• /
• 2003

A method for the automatic segmentation of speech signals is described. The method is dedicated to the construction of a large database for a Text-To-Speech (TTS) synthesis system. The main issue of the work involves the refinement of an initial estimation of phone boundaries which are provided by an alignment, based on a Hidden Market Model(HMM). Multi-layer perceptron (MLP) was used as a phone boundary detector. To increase the performance of segmentation, a technique which individually trains an MLP according to phonetic transition is proposed. The optimum partitioning of the entire phonetic transition space is constructed from the standpoint of minimizing the overall deviation from hand labelling positions. With single speaker stimuli, the experimental results showed that more than 95% of all phone boundaries have a boundary deviation from the reference position smaller than 20 ms, and the refinement of the boundaries reduces the root mean square error by about 25%.

• The Journal of the Acoustical Society of Korea
• /
• v.18 no.8
• /
• pp.94-99
• /
• 1999

It is the hybrid structure of HMM and neural network(NN) that shows high recognition rate in speech recognition algorithms. And it is a method which has majorities of statistical model and neural network model respectively. In this study, we propose a new style of the hybrid structure of semi-continuous HMM(SCHMM) and radial basis function(RBF), which re-estimates weighting coefficients probability affecting observation probability after Baum-Welch estimation. The proposed method takes account of the similarity of basis Auction of RBF's hidden layer and SCHMM's probability density functions so as to discriminate speech signals sensibly through the learned and estimated weighting coefficients of RBF. As simulation results show that the recognition rates of the hybrid structure SCHMM/RBF are higher than those of SCHMM in unlearned speakers' recognition experiment, the proposed method has been proved to be one which has more sensible property in recognition than SCHMM.

• Speech Sciences
• /
• v.1
• /
• pp.153-200
• /
• 1997

In languages such as Japanese, it is very common to observe that short peripheral vowel are completely voiceless when surrounded by voiceless consonants. This phenomenon has been known as Montreal French, Shanghai Chinese, Greek, and Korean. Traditionally this phenomenon has been described as a phonological rule that either categorically deletes the vowel or changes the [+voice] feature of the vowel to [-voice]. This analysis was supported by Sawashima (1971) and Hirose (1971)'s observation that there are two distinct EMG patterns for voiced and devoiced vowel in Japanese. Close examination of the phonetic evidence based on acoustic data, however, shows that these phonological characterizations are not tenable (Jun & Beckman 1993, 1994). In this paper, we examined the vowel devoicing phenomenon in Korean using data from ENG fiberscopic and acoustic recorders of 100 sentences produced by one Korean speaker. The results show that there is variability in the 'degree of devoicing' in both acoustic and EMG signals, and in the patterns of glottal closing and opening across different devoiced tokens. There seems to be no categorical difference between devoiced and voiced tokens, for either EMG activity events or glottal patterns. All of these observations support the notion that vowel devoicing in Korean can not be described as the result of the application of a phonological rule. Rather, devoicing seems to be a highly variable 'phonetic' process, a more or less subtle variation in the specification of such phonetic metrics as degree and timing of glottal opening, or of associated subglottal pressure or intra-oral airflow associated with concurrent tone and stricture specifications. Some of token-pair comparisons are amenable to an explanation in terms of gestural overlap and undershoot. However, the effect of gestural timing on vocal fold state seems to be a highly nonlinear function of the interaction among specifications for the relative timing of glottal adduction and abduction gestures, of the amplitudes of the overlapped gestures, of aerodynamic conditions created by concurrent oral tonal gestures, and so on. In summary, to understand devoicing, it will be necessary to examine its effect on phonetic representation of events in many parts of the vocal tracts, and at many stages of the speech chain between the motor intent and the acoustic signal that reaches the hearer's ear.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.37 no.5A
• /
• pp.388-394
• /
• 2012

In this paper, the influence analysis of food on human body organs is proposed by applying speech signal processing techniques. Until these days, most of researches regarding the influence of food on body organs are such that "A" ingredient of food may produce a good effect on "B" organ. However, the numerical and quantified researches regarding these effects hardly have been performed. This paper therefore proposes a method to quantify the effects by using numerical data, so as to retrieve new facts and informations. Especially, this paper investigates the effect of tomatoes on human heart function. The experiment collects samples of voice signals, before and after 5 minutes, 30 minutes and 1 hour, from 15 males in their 20s who have not abnormal heart function; the voice signal components are applied to measure changes of heart conditions to digitize and quantify the effects of tomatoes on cardiac function.

• The Journal of the Acoustical Society of Korea
• /
• v.25 no.2
• /
• pp.78-87
• /
• 2006

In this paper. we propose an effective mask estimation scheme for missing-feature reconstruction in order to achieve robust speech recognition under unknown noise environments. In the previous work. colored noise is used for training the mask classifer, which is generated from the entire frequency Partitioned signals. However it gives a limited performance under the restricted number of training database. To reflect the spectral events of more various background noise and improve the performance simultaneously. a new Bayesian classifier for mask estimation is proposed, which works independent of other frequency bands. In the proposed method, we employ the colored noise which is obtained by combining colored noises generated from each frequency band in order to reflect more various noise environments and mitigate the 'sparse' database problem. Combined with the cluster-based missing-feature reconstruction. the performance of the proposed method is evaluated on a task of noisy speech recognition. The results show that the proposed method has improved performance compared to the Previous method under white noise. car noise and background music conditions.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.19 no.4
• /
• pp.685-692
• /
• 2024

In this paper, we propose a new CNN model to improve the recognition performance of command voice signals. This method obtains a spectrogram image after performing a short-time Fourier transform (STFT) of the input signal and improves command recognition performance through supervised learning using a CNN model. After Fourier transforming the input signal for each short-time section, a spectrogram image is obtained and multi-classification learning is performed using a CNN deep learning model. This effectively classifies commands by converting the time domain voice signal to the frequency domain to express the characteristics well and performing deep learning training using the spectrogram image for the conversion parameters. To verify the performance of the speech recognition system proposed in this study, a simulation program using Tensorflow and Keras libraries was created and a simulation experiment was performed. As a result of the experiment, it was confirmed that an accuracy of 92.5% could be obtained using the proposed deep learning algorithm.

• The Journal of the Acoustical Society of Korea
• /
• v.14 no.4
• /
• pp.38-45
• /
• 1995

In this paper a way to improve the performance of the long term prediction is proposed, which adopts the Multi-band Excitation (MBE) method in addition to the Code-Excited Linear Prediction (CELP) method at low bit rates below 4.8 kbps. In the proposed method, the multiband long term prediction is performed on the periodic components which still remain after the long term prediction of the conventional CELP method. At this point, the whole frequency region is divided into subbands whose size is equal to the spacing between the harmonics of the fundamental frequency, and the periodic multiband excitation signals. are represented as the sum of sine waves approximately as large as the spectrum of the excitation signals, so that the actual characteristics of the excitation signals can be better taken into account. To evaluate the performance of the proposed method, computer simulation is performed at 4.8 kbps. The 4.8 kbps DoD CELP and the 4.4 kbps IMBE were chosen as the reference vocoders for the speech quality measure. The result of the perceptual speech quality measure showed that the performance of the proposed method is better than that of the 4.8 kbps DoD CELP vocoder, and similar to that of the 4.4 kbps IMBE vocoder.