• The Journal of the Korea Contents Association
• /
• v.6 no.10
• /
• pp.9-16
• /
• 2006

In this paper, the acoustic model output of CSR(Continuous Speech Recognition) was used to analyze news data News database used in this experiment was consisted of 2,093 articles. Due to the low efficiency of language model, conventional Korean CSR is not appropriate to the analysis of news data. This problem could be handled successfully by introducing post-processing work of recognition result of acoustic model. The acoustic model more robust than language model in Korean environment. The result of post-processing work was made into KIF(Keyword information file). When threshold of acoustic model's output level was 100, 86.9% of whole target morpheme was included in post-processing result. At the same condition, applying length information based normalization, 81.25% of whole target morpheme was recognized. The purpose of normalization was to compensate long-length morpheme. According to experiment result, 75.13% of whole target morpheme was recognized KIF(314MB) had been produced from original news data(5,040MB). The decrease rate of absolute information met was approximately 93.8%.

• The Journal of the Acoustical Society of Korea
• /
• v.16 no.4
• /
• pp.57-63
• /
• 1997

In this paper, we study the efficient feature vector extraction method and front-end processing to improve the performance of the speech recognition system using KT(Korea Telecommunication) database collected through various telephone channels. First of all, we compare the recognition performances of the feature vectors known to be robust to noise and environmental variation and verify the performance enhancement of the recognition system using weighted cepstral distance measure methods. The experiment result shows that the recognition rate is increasedby using both PLP(Perceptual Linear Prediction) and MFCC(Mel Frequency Cepstral Coefficient) in comparison with LPC cepstrum used in KT recognition system. In cepstral distance measure, the weighted cepstral distance measure functions such as RPS(Root Power Sums) and BPL(Band-Pass Lifter) help the recognition enhancement. The application of the spectral subtraction method decrease the recognition rate because of the effect of distortion. However, RASTA(RelAtive SpecTrAl) processing, CMS(Cepstral Mean Subtraction) and SBR(Signal Bias Removal) enhance the recognition performance. Especially, the CMS method is simple but shows high recognition enhancement. Finally, the performances of the modified methods for the real-time implementation of CMS are compared and the improved method is suggested to prevent the performance degradation.

• Proceedings of the Acoustical Society of Korea Conference
• /
• spring
• /
• pp.115-118
• /
• 2002

Parameters used in a speaker recognition system are desirable expressing speaker's characteristics filly and have in a speech. That is to say, if inter-speaker than intra-speaker variance a big characteristic, it is useful to distinguish between speakers. Also, to make minimum error between speakers, it is required the improved recognition technology as well as the distinguishing characteristics. When we see the result of recent simulation performance, we obtain more exact performance by using dynamic characteristics and constant characteristics by a speaking habit. Therefore we suggest it to solve this problem as followings. The prosodic information is used by a characteristic vector of speech. Characteristics vector generally using in speaker recognition system is a modeling spectrum information and is working for a high performance in non-noise circumstance. However, it is found a problem that characteristic vector is distorted in noise circumstance and it makes a reduction of recognition rate. In this paper, we change pitch line divided by segment which can estimate a dynamic characteristic and it is used as a recognition characteristic. we confirmed that the dynamic characteristic is very robust in noise circumstance with a simulation. We make a decision of acceptance or rejection by comparing test pattern and recognition rate using the proposed algorithm has more improvement than using spectrum and prosodic information. Especially stational recognition rate can be obtained in noise circumstance through the simulation.

• Journal of the Korean Society of Industry Convergence
• /
• v.21 no.2
• /
• pp.87-93
• /
• 2018

This Study Propose A New Approach To Impliment A Intelligent Robot Control Based on Voice Recognition For Smart Factory Automation Since human usually communicate each other by voices, it is very convenient if voice is used to command humanoid robots or the other type robot system. A lot of researches has been performed about voice recognition systems for this purpose. Hidden Markov Model is a robust statistical methodology for efficient voice recognition in noise environments. It has being tested in a wide range of applications. A prediction approach traditionally applied for the text compression and coding, Prediction by Partial Matching which is a finite-context statistical modeling technique and can predict the next characters based on the context, has shown a great potential in developing novel solutions to several language modeling problems in speech recognition. It was illustrated the reliability of voice recognition by experiments for humanoid robot with 26 joints as the purpose of application to the manufacturing process.

• Phonetics and Speech Sciences
• /
• v.2 no.4
• /
• pp.83-92
• /
• 2010

The present study examined how listeners of Seoul Korean would recover deleted phonemes in consonant cluster simplification. In a phoneme monitoring experiment, listeners had to monitor for C2 (/k/ or /p/) in C1C2C3 when C2 was deleted (C1 was preserved) or preserved (C1 was deleted). The target consonant (C2) was either /k/ or /p/ (e.g., i$\b{lk}$-t${\partial}$lato vs. pa$\b{lp}$-t${\partial}$lato), and there were two listener groups, one group tested in 2002 and the other in 2009. Some points have emerged from the results. First, listeners were able to detect deleted phonemes as accurately and rapidly as preserved phonemes, showing that the physical presence of the acoustic information did not improve the listeners' performance. This suggests that listeners must have relied on language-specific phonological knowledge about the consonant cluster simplification, rather than relying on the low-level acoustic-phonetic information. Second, listener groups (participants in 2002 vs. 2009), differed in processing /p/ versus /k/: listeners in 2009 failed to detect /p/ more frequently than those in 2002, suggesting that the way the consonant cluster sequence is produced and perceived has changed over time. This result was interpreted as coming from statistical patterns of speech production in contemporary Seoul Korean as reported in a recent study by Cho & Kim (2009): /p/ is deleted far more often than /p/ is preserved, which is likely reflected in the way listeners process simplified variants. Finally, listeners processed /k/ more efficiently than /p/, especially when the target was physically present (in C-preserved condition), indicating that listeners benefited more from the presence of /k/ than of /p/. This was interpreted as supporting the view that velars are perceptually more robust than labials, which constrains shaping phonological patterns of the language. These results were then discussed in terms of their implications for theories of spoken word recognition.

• Journal of the Korea Society of Computer and Information
• /
• v.10 no.6 s.38
• /
• pp.47-54
• /
• 2005

In the most of speaker recognition system, speaker's characteristics is extracted from acoustic parameter by speech analysis and we make speaker's reference pattern. Parameters used in speaker recognition system are desirable expressing speaker's characteristics fully and being a few difference whenever it is spoken. Therefore we su99est following to solve this problem. This paper is proposed to use strong spectrum characteristic in non-noise circumstance and prosodic information in noise circumstance. In a stage of making code book, we make the number of data we need to combine spectrum characteristic and Prosodic information. We decide acceptance or rejection comparing test pattern and each model distance. As a result, we obtained more improved recognition rate than we use spectrum and prosodic information especially we obtained stational recognition rate in noise circumstance.

• Journal of the Institute of Electronics Engineers of Korea CI
• /
• v.40 no.5
• /
• pp.289-297
• /
• 2003

The performance of a bimodal system is affected by the accuracy of the endpoint detection from the input signal as well as the performance of the speech recognition or lipreading system. In this paper, we propose the endpoint detection method which detects the endpoints from the audio and video signal respectively and utilizes the signal to-noise ratio (SNR) estimated from the input audio signal to select the reliable endpoints to the acoustic noise. In other words, the endpoints are detected from the audio signal under the high SNR and from the video signal under the low SNR. Experimental results show that the bimodal system using the proposed endpoint detector achieves satisfactory recognition rates, especially when the acoustic environment is quite noisy.

• Proceedings of the Korean Information Science Society Conference
• /
• 2005.11b
• /
• pp.733-735
• /
• 2005

최근 음성인식 분야 에서는 잡음 환경에서 좀 더 신뢰도 높은 음성 인식 결과물 얻기 위하여 인식 결과 도출 단계에서 여러 가지 정보를 융합 하는 방법이나 인식결과를 후처리 하여 새로운 결과를 얻어 내는 방법들이 연구 되고 있다. 본 논문에서는 개인 모바일 기기에서의 음성 인식 환경에서 사용자의 발화 패턴 정보를 가지는 문맥 정보를 활용함으로서 잡음 환경에서의 음성 정보 손실에 따른 인식률 하락을 보완하는 방법을 제안한다. 먼저 사용자의 기기 사용 로그나 발화 로그 정보로부터 특정 명령어들의 순차적 발화 패턴을 마이닝하여 문맥 정보를 구성한다. 이 후 음성 발화시에 인식기의 최종 인식 결과에 대한 신뢰도가 떨어진다고 판단될 때 앞서 얻어진 문맥 정보의 신뢰도를 인식기의 각 후보단어들의 인식률과 융합하여 새로운 인식 결과를 도출해 낸다. 이러한 과정에서 인식기 결과에 대한 신뢰성을 판단하는 기준을 실험을 통하여 결정 하였으며 신뢰성이 기준 이하일 경우의 융합 과정을 위하여 후보 단어 인식률과 문맥정보를 적절히 융합할 수 있는 방법을 제안한다.

• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 2007.11a
• /
• pp.431-434
• /
• 2007

본 논문에서는 인간의 감정 변화의 영향을 적게 받는 음성 인식 시스템의 특정 파라메터에 관한 연구를 수행하였다. 이를 위하여 우선 다양한 감정이 포함된 음성 데이터베이스를 사용하여 감정 변화가 음성 인식 시스템의 성능에 미치는 영향과 감정 변화의 영향을 적게 받는 특정 파라메터에 관한 연구를 수행하였다. 본 연구에서는 LPC 켑스트럼 계수, 멜 켑스트럼 계수, 루트 켑스트럼 계수, PLP 계수와 RASTA 처리를 한 멜 켑스트럼 계수와 음성의 에너지를 사용하였다. 또한 음성에 포함된 편의(bias)를 제거하는 방법으로 CMS 와 SBR 방법을 사용하여 그 성능을 비교하였다. HMM 기반의 화자독립 단어 인식기를 사용한 실험 결과에서 RASTA 멜 켑스트럼과 델타 켑스트럼을 사용하고 신호편의 제거 방법으로 CMS를 사용한 경우에 가장 우수한 성능을 나타내었다. 이러한 것은 멜 켑스트럼을 사용한 기준 시스템과 비교하여 59%정도 오차가 감소된 것이다.

• Proceedings of the Korean Information Science Society Conference
• /
• 2004.10b
• /
• pp.580-582
• /
• 2004

스펙트럼 차감과정에서 발생하는 잔류 잡음을 제거하는 방법으로 파라메터를 이용하는 적응 스펙트럼 차감법이 있다. 이는 파라메터를 증가시켜 잔류 잡음을 감소시키는 방법이지만 파라메터를 과도하게 증가시킬 경우 음성 왜곡이 발생한다. 따라서, 적절한 파라메터를 추출하기 위하여 SNR이나, 마스킹 효과 등을 이용한 방법들이 제안되었으나 과도한 잡음의 제거로 인한 음성 왜곡 문제와 낮은 SNR에서 부정확한 파라메터의 추출 문제는 여전히 해결해야 할 과제로 남아있다. 본 논문은 기존의 SNR을 이용한 방법에 마스킹 효과를 적용한 수정된 적응 스펙트럼 차감법을 제안한다. 제안된 방법에서는 마스킹 임계치를 이용하여 잡음 추정값을 재 계산 항으로써 SNR을 향상시켰고, 이를 이용하여 파라메터를 추출함으로써 성능을 개선했다 성능평가 결과, 제안한 차감법을 적용한 음성신호를 고립단어 음성인식 시스템에 적용했을 때 기존의 방법 보다 인식률이 향상된 것을 확인할 수 있었다.