• ETRI Journal
• /
• v.30 no.6
• /
• pp.865-867
• /
• 2008

A conventional environment adaptation for robust speech recognition is usually conducted using transform-based techniques. Here, we present a discriminative adaptation strategy based on a multi-condition-trained model, and propose a new method to provide universal application to a new environment using the environment's specific conditions. Experimental results show that a speech recognition system adapted using the proposed method works successfully for other conditions as well as for those of the new environment.

• Proceedings of the IEEK Conference
• /
• 2007.07a
• /
• pp.423-424
• /
• 2007

In this paper, we propose a scalable high-quality speech reconstruction method for distributed speech recognition (DSR). It is difficult to reconstruct speech of high quality with MFCCs at the DSR server. Depending on the bit-rate available by the DSR system, we can send additional information associated with speech coding to the DSR sorrel, where the bit-rate is variable from 4.8 kbit/s to 11.4 kbit/s. The experimental results show that the speech quality reproduced by the proposed method when the bit-rate is 11.4 kbit/s is comparable with that of ITU-T G.729 under both ideal channel and frame error channel conditions while the performance of DSR is maintained to that of wireline speech recognition.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.16 no.3
• /
• pp.465-470
• /
• 2021

A robust speech recognition technology is required that does not degrade the performance of speech recognition and the quality of the speech when speech recognition is performed in an actual environment of the speech mixed with noise. With the development of such speech recognition technology, it is necessary to develop an application that achieves stable and high speech recognition rate even in a noisy environment similar to the human speech spectrum. Therefore, this paper proposes a speech enhancement algorithm that processes a noise suppression based on the MMSA-STSA estimation algorithm, which is a short-time spectral amplitude method based on the error of the least mean square. This algorithm is an effective nonlinear speech enhancement algorithm based on a single channel input and has high noise suppression performance. Moreover this algorithm is a technique that reduces the amount of distortion of the speech based on the statistical model of the speech. In this experiment, in order to verify the effectiveness of the MMSA-STSA estimation algorithm, the effectiveness of the proposed algorithm is verified by comparing the input speech waveform and the output speech waveform.

• MALSORI
• /
• no.43
• /
• pp.45-55
• /
• 2002

High reliability of automatic speaker recognition regardless of voice conditions is necessary for forensic application. Audio recordings in real cases are not consistent in voice conditions, such as duration, time interval of recording, given text or conversational speech, transmission channel, etc. In this study the variation of verification error rate of ASR system with the voice conditions was investigated. As a result in order to decrease both false rejection rate and false acception rate, the various voices should be used for training and the duration of train voices should be longer than the test voices.

• Proceedings of the KSPS conference
• /
• 2006.05a
• /
• pp.37-40
• /
• 2006

In this paper, we propose a VQ codebook design of speech recognition feature parameters in order to improve the performance of a distributed speech recognition system. For the context-dependent HMMs, a VQ codebook should be correlated with phonetic distributions in the training data for HMMs. Thus, we focus on a selection method of training data based on phonetic distribution instead of using all the training data for an efficient VQ codebook design. From the speech recognition experiments using the Aurora 4 database, the distributed speech recognition system employing a VQ codebook designed by the proposed method reduced the word error rate (WER) by 10% when compared with that using a VQ codebook trained with the whole training data.

• International journal of advanced smart convergence
• /
• v.10 no.2
• /
• pp.79-85
• /
• 2021

Speech recognition technology refers to a technology in which a computer interprets the speech language spoken by a person and converts the contents into text data. This technology has recently been combined with artificial intelligence and has been used in various fields such as smartphones, set-top boxes, and smart TVs. Examples include Google Assistant, Google Home, Samsung's Bixby, Apple's Siri and SK's NUGU. Google and Daum Kakao offer free open APIs for speech recognition technologies. This paper selects three APIs that are free to use by ordinary users, and compares each recognition rate according to the three types. First, the recognition rate of "numbers" and secondly, the recognition rate of "Ga Na Da Hangul" are conducted, and finally, the experiment is conducted with the complete sentence that the author uses the most. All experiments use real voice as input through a computer microphone. Through the three experiments and results, we hope that the general public will be able to identify differences in recognition rates according to the applications currently available, helping to select APIs suitable for specific application purposes.

• MALSORI
• /
• no.46
• /
• pp.87-102
• /
• 2003

In this paper, we present an English pronunciation correction system for Korean speakers and show some of experimental results on it. The aim of the system is to detect mispronounced phonemes in spoken words and to give appropriate correction comments to users. There are several English pronunciation correction systems adopting speech recognition technology, however, most of them use conventional speech recognition engines. From this reason, they could not give phoneme based correction comments to users. In our system, we build two kinds of phoneme models: standard native speaker models and Korean's error models. We also design recognition network based on phonemes to detect Koreans' common mispronunciations. We get 90% detection rate in insertion/deletion/replacement of phonemes, but we cannot get high detection rate in diphthong split and accents.

• Phonetics and Speech Sciences
• /
• v.9 no.2
• /
• pp.103-110
• /
• 2017

The pole filtering concept has been successfully applied to cepstral feature normalization techniques for noise-robust speech recognition. In this paper, it is proposed to apply the pole filtering selectively only to the speech intervals, in order to further improve the recognition performance for short utterances in noisy environments. Experimental results on AURORA 2 task with clean-condition training show that the proposed selectively pole-filtered cepstral mean normalization (SPFCMN) and selectively pole-filtered cepstral mean and variance normalization (SPFCMVN) yield error rate reduction of 38.6% and 45.8%, respectively, compared to the baseline system.

• MALSORI
• /
• no.60
• /
• pp.145-164
• /
• 2006

This paper proposes class histogram equalization (CHEQ) to compensate noisy acoustic features for robust speech recognition. CHEQ aims to compensate for the acoustic mismatch between training and test speech recognition environments as well as to reduce the limitations of the conventional histogram equalization (HEQ). In contrast to HEQ, CHEQ adopts multiple class-specific distribution functions for training and test environments and equalizes the features by using their class-specific training and test distributions. According to the class-information extraction methods, CHEQ is further classified into two forms such as hard-CHEQ based on vector quantization and soft-CHEQ using the Gaussian mixture model. Experiments on the Aurora 2 database confirmed the effectiveness of CHEQ by producing a relative word error reduction of 61.17% over the baseline met-cepstral features and that of 19.62% over the conventional HEQ.

• Proceedings of the KSPS conference
• /
• 2005.11a
• /
• pp.103-106
• /
• 2005

In this paper, we propose a language modeling approach to improve the performance of a large vocabulary continuous speech recognition system. The proposed approach is based on the active learning framework that helps to select a text corpus from a plenty amount of text data required for language modeling. The perplexity is used as a measure for the corpus selection in the active learning. From the recognition experiments on the task of continuous Korean speech, the speech recognition system employing the language model by the proposed language modeling approach reduces the word error rate by about 6.6 % with less computational complexity than that using a language model constructed with randomly selected texts.