• MALSORI
• /
• v.55
• /
• pp.103-118
• /
• 2005

In this paper, we describe a morpheme-based pronunciation lexicon useful for Korean LVCSR. The phonemic-context-dependent multiple pronunciation lexicon improves the recognition accuracy when cross-morpheme pronunciation variations are distinguished from within-morpheme pronunciation variations. Since adding all possible pronunciation variants to the lexicon increases the lexicon size and confusability between lexical entries, we have developed a lexicon pruning scheme for optimal selection of pronunciation variants to improve the performance of Korean LVCSR. By building a proposed pronunciation lexicon, an absolute reduction of $0.56\%$ in WER from the baseline performance of $27.39\%$ WER is achieved by cross-morpheme pronunciation variations model with a phonemic-context-dependent multiple pronunciation lexicon. On the best performance, an additional reduction of the lexicon size by $5.36\%$ is achieved from the same lexical entries.

• Proceedings of the Korean Information Science Society Conference
• /
• 2011.06b
• /
• pp.231-234
• /
• 2011

Audiovisual-based human emotion recognition can be considered a good approach for multimodal humancomputer interaction. However, the optimal multimodal information fusion remains challenges. In order to overcome the limitations and bring robustness to the interface, we propose a framework of automatic human emotion recognition system from speech and face display. In this paper, we develop a new approach for fusing information in model-level based on the relationship between speech and face expression to detect automatic temporal segments and perform multimodal information fusion.

• Proceedings of the Korean Society for Emotion and Sensibility Conference
• /
• 2000.04a
• /
• pp.366-369
• /
• 2000

This paper examines three algorithms to recognize speaker's emotion using the speech signals. Target emotions are happiness, sadness, anger, fear, boredom and neutral state. MLB(Maximum-Likeligood Bayes), NN(Nearest Neighbor) and HMM (Hidden Markov Model) algorithms are used as the pattern matching techniques. In all cases, pitch and energy are used as the features. The feature vectors for MLB and NN are composed of pitch mean, pitch standard deviation, energy mean, energy standard deviation, etc. For HMM, vectors of delta pitch with delta-delta pitch and delta energy with delta-delta energy are used. We recorded a corpus of emotional speech data and performed the subjective evaluation for the data. The subjective recognition result was 56% and was compared with the classifiers' recognition rates. MLB, NN, and HMM classifiers achieved recognition rates of 68.9%, 69.3% and 89.1% respectively, for the speaker dependent, and context-independent classification.

• International Journal of Computer Science & Network Security
• /
• v.23 no.4
• /
• pp.48-54
• /
• 2023

Nowadays, a blind man finds it very difficult to cross the roads. They should be very vigilant with every step they take. To resolve this problem, Convolutional Neural Networks(CNN) is a best method to analyse the data and automate the model without intervention of human being. In this work, a traffic signal recognition system is designed using CNN for the visually impaired. To provide a safe walking environment, a voice message is given according to light state and timer state at that instance. The developed model consists of two phases, in the first phase the CNN model is trained to classify different images captured from traffic signals. Common Objects in Context (COCO) labelled dataset is used, which includes images of different classes like traffic lights, bicycles, cars etc. The traffic light object will be detected using this labelled dataset with help of object detection model. The CNN model detects the color of the traffic light and timer displayed on the traffic image. In the second phase, from the detected color of the light and timer value a text message is generated and sent to the text-to-speech conversion model to make voice guidance for the blind person. The developed traffic light recognition model recognizes traffic light color and countdown timer displayed on the signal for safe signal crossing. The countdown timer displayed on the signal was not considered in existing models which is very useful. The proposed model has given accurate results in different scenarios when compared to other models.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.31 no.9C
• /
• pp.853-858
• /
• 2006

TMHMM(Tied Mixture Hidden Markov Model) is an important approach to reduce the number of free parameters in speech recognition. However, this model suffers from a degradation in recognition accuracy due to its GPDF (Gaussian Probability Density Function) clustering error. This paper proposes a clustering algorithm, called HCNN(Homogeneous Centroid Neural network), to cluster acoustic feature vectors in TMHMM. Moreover, the HCNN uses the heterogeneous distance measure to allocate more code vectors in the heterogeneous areas where probability densities of different states overlap each other. When applied to Korean digit isolated word recognition, the HCNN reduces the error rate by 9.39% over CNN clustering, and 14.63% over the traditional K-means clustering.

• The Korean Journal of Applied Statistics
• /
• v.32 no.2
• /
• pp.213-227
• /
• 2019

In this paper, we propose an end-to-end deep learning model combining Bayesian neural network with Korean speech recognition. In the past, Korean speech recognition was a complicated task due to the excessive parameters of many intermediate steps and needs for Korean expertise knowledge. Fortunately, Korean speech recognition becomes manageable with the aid of recent breakthroughs in "End-to-end" model. The end-to-end model decodes mel-frequency cepstral coefficients directly as text without any intermediate processes. Especially, Connectionist Temporal Classification loss and Attention based model are a kind of the end-to-end. In addition, we combine Bayesian neural network to implement the end-to-end model and obtain Monte Carlo estimates. Finally, we carry out our experiments on the "WorimalSam" online dictionary dataset. We obtain 4.58% Word Error Rate showing improved results compared to Google and Naver API.

• Proceedings of the IEEK Conference
• /
• 2003.07e
• /
• pp.2220-2223
• /
• 2003

In this paper we used Support Vector Machines(SVMs) recently proposed as the loaming method, one of Artificial Neural Network, to divide continuous speech into phonemes, an initial, medial, and final sound, and then, performed continuous speech recognition from it. Decision boundary of phoneme is determined by algorithm with maximum frequency in a short interval. Recognition process is performed by Continuous Hidden Markov Model(CHMM), and we compared it with another phoneme divided by eye-measurement. From experiment we confirmed that the method, SVMs, we proposed is more effective in an initial sound than Gaussian Mixture Models(GMMs).

• Proceedings of the IEEK Conference
• /
• 2003.07e
• /
• pp.2240-2243
• /
• 2003

In this paper. we present a new speech recognition system using DSP chip. DSP chip used TMS320c6711 of TI. We designed hardware system including acoustic model, word list and code book in flash memory. The word candidates are recognized based on CV, VCCV, and VC units HMM. This system can be applied to various electric ＆ electronic devices: home automation, robotics etc.

• Proceedings of the Acoustical Society of Korea Conference
• /
• 1994.06a
• /
• pp.1082-1087
• /
• 1994

A set of text database is indispensable to the probabilistic models for speech recognition, linguistic model, and machine translation. We introduce an environment to canstruct text databases : an automatic tagging system and a set of tools for lexical knowledge acquisition, which provides the facilities of automatic part of speech recognition and guessing.

• The Journal of the Acoustical Society of Korea
• /
• v.18 no.8
• /
• pp.3-8
• /
• 1999

Previous variable vocabulary speech recognition systems with context-independent acoustic modeling, could not represent the effect of neighboring phonemes. To solve this problem, we use allophone-based context-dependent acoustic model. This paper describes the method to improve acoustic model of the system effectively. Acoustic model is improved by using allophone clustering technique that uses entropy as a similarity measure and the optimal allophone model is generated by changing the number of allophones. We evaluate performance of the improved system by using Phonetically Optimized Words(POW) DB and PC commands(PC) DB. As a result, the allophone model composed of six hundreds allophones improved the recognition rate by 13% from the original context independent model m POW test DB.