• The Journal of the Acoustical Society of Korea
• /
• v.22 no.1E
• /
• pp.3-10
• /
• 2003

This paper describes two kinds of neural networks for stereoscopic vision, which have been applied to an identification of human speech. In speech recognition based on the stereoscopic vision neural networks (SVNN), the similarities are first obtained by comparing input vocal signals with standard models. They are then given to a dynamic process in which both competitive and cooperative processes are conducted among neighboring similarities. Through the dynamic processes, only one winner neuron is finally detected. In a comparative study, with, the average phoneme recognition accuracy on the two-layered SVNN was 7.7% higher than the Hidden Markov Model (HMM) recognizer with the structure of a single mixture and three states, and the three-layered was 6.6% higher. Therefore, it was noticed that SVNN outperformed the existing HMM recognizer in phoneme recognition.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.18 no.6
• /
• pp.1692-1705
• /
• 2024

In this paper, we present a method that integrates a Grammar Transducer as an external language model to enhance the accuracy of the pre-trained Korean End-to-end (E2E) Automatic Speech Recognition (ASR) model. The E2E ASR model utilizes the Connectionist Temporal Classification (CTC) loss function to derive hypothesis sentences from input audio. However, this method reveals a limitation inherent in the CTC approach, as it fails to capture language information from transcript data directly. To overcome this limitation, we propose a fusion approach that combines a clause-level n-gram language model, transformed into a Weighted Finite-State Transducer (WFST), with the E2E ASR model. This approach enhances the model's accuracy and allows for domain adaptation using just additional text data, avoiding the need for further intensive training of the extensive pre-trained ASR model. This is particularly advantageous for Korean, characterized as a low-resource language, which confronts a significant challenge due to limited resources of speech data and available ASR models. Initially, we validate the efficacy of training the n-gram model at the clause-level by contrasting its inference accuracy with that of the E2E ASR model when merged with language models trained on smaller lexical units. We then demonstrate that our approach achieves enhanced domain adaptation accuracy compared to Shallow Fusion, a previously devised method for merging an external language model with an E2E ASR model without necessitating additional training.

• MALSORI
• /
• no.46
• /
• pp.103-115
• /
• 2003

In this paper, we proposed Background Model Set algorithm used in the speaker verification to improve calculating confidence measure(CM) in speech recognition. CM is to display relative likelihood between recognized models and antiphone models. In previous method calculating of CM, we calculated probability and standard deviation using all phonemes in composition of antiphone models. At this process, antiphone CM brought bad recognition result. Also, recognition time increases. In order to solve this problem, we studied about method to reconstitute average and standard deviation using BMS algorithm in CM calculation.

• The Journal of The Korea Institute of Intelligent Transport Systems
• /
• v.5 no.2 s.10
• /
• pp.1-10
• /
• 2006

In this paper, VTS(Vector Taylor Series) algorithm, which was proposed by Moreno at Carnegie Mellon University in 1996, is analyzed and simulated. VTS is considered to be one of the robust speech recognition techniques where model parameter conversion technique is adapted. To evaluation performance of the VTS algorithm, We used CMN(Cepstral Mean Normalization) technique which is one of the well-known noise processing methods. And the recognition rate is evaluated when white gaussian and street noise are employed as background noise. Also, the simulation result is analyzed in order to be compared with the previous one which was performed by Moreno.

• Journal of KIISE:Computing Practices and Letters
• /
• v.13 no.5
• /
• pp.338-342
• /
• 2007

In this paper we propose a user adaptive post-processing method to improve the accuracy of speaker dependent, isolated word speech recognition, particularly for mobile devices. Our method considers the recognition result of the basic recognizer simply as a high-level speech feature and processes it further for correct recognition result. Our method learns correlation between the output of the basic recognizer and the correct final results and uses it to correct the erroneous output of the basic recognizer. A multi-layer perceptron model is built for each incorrectly recognized word with high frequency. As the result of experiments, we achieved a significant improvement of 41% in recognition accuracy (41% error correction rate).

• The Journal of the Acoustical Society of Korea
• /
• v.21 no.3
• /
• pp.323-329
• /
• 2002

This paper proposes a frame reliability weighting method to compensate for a time-selective noise that occurs at random positions of speech signal contaminating certain parts of the speech signal. Speech frames have different degrees of reliability and the reliability is proportional to SNR (signal-to noise ratio). While it is feasible to estimate frame Sl? by using the noise information from non-speech interval under a stationary noisy situation, it is difficult to obtain noise spectrum for a time-selective noise. Therefore, we used statistical models of clean speech for the estimation of the frame reliability. The proposed MFR (model-based frame reliability) approximates frame SNR values using filterbank energy vectors that are obtained by the inverse transformation of input MFCC (mal-frequency cepstral coefficient) vectors and mean vectors of a reference model. Experiments on various burnt noises revealed that the proposed method could represent the frame reliability effectively. We could improve the recognition performance by using MFR values as weighting factors at the likelihood calculation step.

• The Journal of the Acoustical Society of Korea
• /
• v.38 no.1
• /
• pp.47-50
• /
• 2019

This paper proposes an acoustic model adaptation method for effective speech recognition in noisy environments. In the proposed algorithm, the noise corruption function is estimated employing DNN (Deep Neural Network), and the function is applied to the model parameter estimation. The experimental results using the Aurora 2.0 framework and database demonstrate that the proposed model adaptation method shows more effective in known and unknown noisy environments compared to the conventional methods. In particular, the experiments of the unknown environments show 15.87 % of relative improvement in the average of WER (Word Error Rate).

• 전자공학회논문지 IE
• /
• v.45 no.3
• /
• pp.20-26
• /
• 2008

Two independent component analysis(ICA) algorithms were applied for the recognition of speech signals corrupted by a car engine noise. Speech recognition was performed by hidden markov model(HMM) for the estimated signals and recognition rates were compared with those of orginal speech signals which are not corrupted. Two different ICA methods were applied for the estimation of speech signals, one of which is FastICA algorithm that maximizes negentropy, the other is information-maximization approach that maximizes the mutual information between inputs and outputs to give maximum independence among outputs. Word recognition rate for the Korean news sentences spoken by a male anchor is 87.85%, while there is 1.65% drop of performance on the average for the estimated speech signals by FastICA and 2.02% by information-maximization for the various signal to noise ratio(SNR). There is little difference between the methods.

• Proceedings of the IEEK Conference
• /
• 1998.10a
• /
• pp.1255-1258
• /
• 1998

Hidden Markov Model (HMM) is the most widely used method in speech recognition. In general, HMM parameters are trained to have maximum likelihood (ML) for training data. This method doesn't take account of discrimination to other words. To complement this problem, this paper proposes a word verification method by re-recognition of the recognized word and its similar word using the discriminative function between two words. The similar word is selected by calculating the probability of other words to each HMM. The recognizer haveing discrimination to each word is realized using the weighting to each state and the weighting is calculated by genetic algorithm.

• Proceedings of the IEEK Conference
• /
• 1999.06a
• /
• pp.671-674
• /
• 1999

We report the new method for speaker recognition. Until now, many researchers have used HMM (Hidden Markov Model) with cepstral coefficient or neural network for speaker recognition. Here, we introduce the method of speaker recognition using eigenspace. This method can reduce the training and recognition time of speaker recognition system. In proposed method, we use the low rank model of the speech eigenspace. In experiment, we obtain good recognition result.