• The Journal of the Acoustical Society of Korea
• /
• 제16권2E호
• /
• pp.32-37
• /
• 1997

This paper proposes a isolated speech recognition method of Korean digits using a TDNN(Time Delay Neural Network) which is able to recognizc time-varying speech properties. We also make an investigation of effect on network parameter of TDNN ; hidden layers and time-delays. TDNNs in our experiments consist of 2 and 3 hidden layers and have several time-delays. From experiment result, TDNN structure which has 2 hidden-layers, gives a good result for speech recognition of Korean digits. Mis-recognition by time-delays can be improved by changing TDNN structures and mis-recognition separated from time-delays can be improved by changing input patterns.

• Journal of Electrical Engineering and information Science
• /
• 제1권1호
• /
• pp.98-107
• /
• 1996

In this paper, we propose a scheme that maps the time-delay neural network (TDNN) into the neurocomputer called EMIND-II which has the wavefront toroidal mesh-array structure. This neurocomputer is scalable, consists of many timeshared virtual neurons, is equipped with programmable on-chip learning, and is versatile for building many types of neural networks. Also we define the programming model of this array and derive the parallel algorithms about TDNN for the proposed neurocomputer EMIND-II. In addition, the computational complexities for the parallel and serial algorithms are compared. Finally, we introduce an application of this neurocomputer to word recognition.

• 한국음향학회:학술대회논문집
• /
• 한국음향학회 1992년도 학술논문발표회 논문집 제11권 1호
• /
• pp.61-65
• /
• 1992

This paper presents a study on Neural Networks for Phoneme Recognition and performs phoneme recognition using TDNN(Time Delay Neural Network). Also, this paper proposes new training algorithm for speech recognition using neural nets that proper to large scale TDNN. Because phoneme recognition is indispensable for continuous speech recognition, this paper uses TDNN to get accurate recognition result of phoneme. And this paper proposes new training algorithm that can converge TDNN to optimal state regardless of the number of phoneme to be recognized. The result of recognition on three phoneme classes shows recognition rate of 9.1%. And this paper proves that proposed algorithm is a efficient method for high performance and reducing convergence time.

• 한국멀티미디어학회:학술대회논문집
• /
• 한국멀티미디어학회 2001년도 추계학술발표논문집
• /
• pp.85-90
• /
• 2001

본 논문에서는 한국어 숫자음 인식을 위한 시간 지연 신경망(Time delay neural network-TDNN)과 은닉 마르코프 모델(Midden Markov Model-HMM)의 결합 방법에 대해서 연구하였고 그 성능을 측정하였으며, 기존의 시스템과 비교 평가하였다. 이 알고리즘은 TDNN과 HMM의 구조적인 결합에 기반하고 있는데 TDNN의 두번째 은닉층의 출력이 HMM의 입력으로 들어가도록 구성되었다. 그러면 HMM은 TDNN의 출력으로 각 단어에 대해서 훈련과정을 거치게 된다. 이렇게 구성된 인식알고리즘은 TDNN의 뛰어난 단기간(Short-time)분류 기능과 HMM의 시간 정렬(time-warping) 능력을 동시에 갖게 된다. 위의 과정을 컴퓨터 시뮬레이션을 이용하여 구현하였으며, 한사람의 음성을 녹음하여 실험한 결과 기존의 TDNN만으로 만들어진 인식기보다는 3%, HMM만으로 구성된 인식기 보다는 5.7% 나은 성능을 얻을 수 있었다.

• 한국통신학회논문지
• /
• 제21권2호
• /
• pp.534-540
• /
• 1996

This paper presents an Adaptive Time-Delay Recurrent Neural Network (ATRN) for learning and recognition of temporal correlations of temporal patterns. The ATRN employs adaptive time-delays and recurrent connections, which are inspired from neurobiology. In the ATRN, the adaptive time-delays make the ATRN choose the optimal values of time-delays for the temporal location of the important information in the input parrerns, and the recurrent connections enable the network to encode and integrate temporal information of sequences which have arbitrary interval time and arbitrary length of temporal context. The ATRN described in this paper, ATNN proposed by Lin, and TDNN introduced by Waibel were simulated and applied to the chaotic time series preditcion of Mackey-Glass delay-differential equation. The simulation results show that the normalized mean square error (NMSE) of ATRN is 0.0026, while the NMSE values of ATNN and TDNN are 0.014, 0.0117, respectively, and in temporal learning, employing recurrent links in the network is more effective than putting multiple time-delays into the neurons. The best performance is attained bythe ATRN. This ATRN will be sell applicable for temporally continuous domains, such as speech recognition, moving object recognition, motor control, and time-series prediction.

• 대한의용생체공학회:학술대회논문집
• /
• 대한의용생체공학회 1994년도 추계학술대회
• /
• pp.165-168
• /
• 1994

We proposed a new model of neural network, called Time Delay Counter-Propagation Neural network (TDCPN). This model is combined properly by the merits of Time Delay Neural Network (TDNN) structure and those of Counter - Propagation Neural network (CPN) learning rule, so that increase recognition rate but decrease total teaming time. And we use this model to simulate classification of EMG signals, and compare the recognition rate and teaming time with those of another neural network model. As a result of simulation, the proposed model is proved to be very effective.

• 대한의용생체공학회:의공학회지
• /
• 제17권3호
• /
• pp.395-401
• /
• 1996

In this paper a new neural network model, time delayed counterpropagation neural networks (TDCPN) which have high recognition rate and short total learning time, is proposed for electromyogram(EMG) recognition. Signals the proposed model increases the recognition rates after learned the regional temporal correlation of patterns using time delay properties in input layer, and decreases the learning time by using winner-takes-all learning rule. The ouotar learning rule is put at the output layer so that the input pattern is able to map a desired output. We test the performance of this model with EMG signals collected from a normal subject. Experimental results show that the recognition rates of the suggested model is better and the learning time is shorter than those of TDNN and CPN.

• 전자공학회논문지B
• /
• 제32B권1호
• /
• pp.171-179
• /
• 1995

In many continuous speech recognition systems, phoneme is used as a basic recognition unit However, the coarticulation generated among neighboring phonemes makes difficult to recognize phonemes consistently. This paper proposes allophone as an alternative recognition unit. We have classified each phoneme into three different allophone groups by the location of phoneme within a syllable. For a recognition algorithm, time-delay neural network(TDNN) has been designed. To recognize all Korean allophones, TDNNs are constructed in modular fashion according to acoustic-phonetic features (e.g. voiced/unvoiced, the location of phoneme within a word). Each TDNN is trained independently, and then they are integrated hierarchically into a whole speech recognition system. In this study, we have experimented Korean plosives with phoneme-based recognition system and allophone-based recognition system. Experimental results show that allophone-based recognition is much less affected by the coarticulation.

• 한국정보처리학회:학술대회논문집
• /
• 한국정보처리학회 2003년도 춘계학술발표논문집 (상)
• /
• pp.285-288
• /
• 2003

기존의 침입탐지 시스템은 오용탐지모델이 널리 사용되고 있다. 이 모델은 낮은 오판율(False Alarm rates)을 가지고 있으나 새로운 공격에 대해 전문가시스템(Expert Systems)에 의한 규칙추가를 필요로 하고, 그 규칙과 완전히 매칭되는 시그너처만 공격으로 탐지하므로 변형된 공격을 탐지하지 못한다는 문제점을 가지고 있다. 본 논문에서는 이러한 문제점을 보완하기 위해 주성분분석(Principle Component Analysis ; 이하 PCA)과 시간지연신경망(Time Delay Neural Network ; 이하 TDNN)을 이용한 침입탐지 시스템을 제안한다. 패킷은 PCA를 이용하여 주성분을 결정하고 패킷이미지패턴으로 만든다. 이 연속된 패킷이미지패턴을 시간지연신경망의 학습패턴으로 사용한다.

• Environmental Engineering Research
• /
• 제21권4호
• /
• pp.333-340
• /
• 2016

An Artificial Neural Network including a Radial Basis Function (RBF) and a Time Delay Neural Network (TDNN) was used to predict total dissolved solid (TDS) in the river Zayanderud. Water quality parameters in the river for ten years, 2001-2010, were prepared from data monitored by the Isfahan Regional Water Authority. A factor analysis was applied to select the inputs of water quality parameters, which obtained total hardness, bicarbonate, chloride and calcium. Input data to the neural networks were pH, $Na^+$, $Mg^{2+}$, Carbonate ($CO{_3}^{-2}$), $HCO{_3}^{-1}$, $Cl^-$, $Ca^{2+}$ and Total hardness. For learning process 5-fold cross validation were applied. In the best situation, the TDNN contained 2 hidden layers of 15 neurons in each of the layers and the RBF had one hidden layer with 100 neurons. The Mean Squared Error and the Mean Bias Error for the TDNN during the training process were 0.0006 and 0.0603 and for the RBF neural network the mentioned errors were 0.0001 and 0.0006, respectively. In the RBF, the coefficient of determination ($R^2$) and the index of agreement (IA) between the observed data and predicted data were 0.997 and 0.999, respectively. In the TDNN, the $R^2$ and the IA between the actual and predicted data were 0.957 and 0.985, respectively. The results of sensitivity illustrated that $Ca^{2+}$ and $SO{_4}^{2-}$ parameters had the highest effect on the TDS prediction.