• MALSORI
• /
• no.62
• /
• pp.85-96
• /
• 2007

In this paper, we carried out recognition experiments for noisy speech having various levels of car noise and output of an audio system using the speech interface. The speech interface consists of three parts: pre-processing, acoustic echo canceller, post-processing. First, a high pass filter is employed as a pre-processing part to remove some engine noises. Then, an echo canceller implemented by using an FIR-type filter with an NLMS adaptive algorithm is used to remove the music or speech coming from the audio system in a car. As a last part, the MMSE-STSA based speech enhancement method is applied to the out of the echo canceller to remove the residual noise further. For recognition experiments, we generated test signals by adding music to the car noisy speech from Aurora 2 database. The HTK-based continuous HMM system is constructed for a recognition system. Experimental results show that the proposed speech interface is very promising for robust speech recognition in a noisy car environment.

• IEIE Transactions on Smart Processing and Computing
• /
• v.3 no.5
• /
• pp.259-266
• /
• 2014

In this paper, we propose a novel technique for noise robust automatic speech recognition (ASR). The development of ASR techniques has made it possible to recognize isolated words with a near perfect word recognition rate. However, in a highly noisy environment, a distinct mismatch between the trained speech and the test data results in a significantly degraded word recognition rate (WRA). Unlike conventional ASR systems employing Mel-frequency cepstral coefficients (MFCCs) and a hidden Markov model (HMM), this study employ histogram of oriented gradient (HOG) features and a Support Vector Machine (SVM) to ASR tasks to overcome this problem. Our proposed ASR system is less vulnerable to external interference noise, and achieves a higher WRA compared to a conventional ASR system equipped with MFCCs and an HMM. The performance of our proposed ASR system was evaluated using a phonetically balanced word (PBW) set mixed with artificially added noise.

• Proceedings of the IEEK Conference
• /
• 2002.07a
• /
• pp.98-101
• /
• 2002

This paper reports a parallel architecture in a HMM based speech recognition system for a low-power LSI design. The proposed architecture calculates output probability of continuous HMM (CHMM) by using concurrent and pipeline processing. They enable to reduce memory access and have high computing efficiency. The novel point is the efficient use of register arrays that reduce memory access considerably compared with any conventional method. The implemented system can achieve a real time response with lower clock in a middle size vocabulary recognition task (100-1000 words) by using this technique.

• Proceedings of the Acoustical Society of Korea Conference
• /
• 1994.06c
• /
• pp.373-377
• /
• 1994

연속음성 인식 시스템 구성을 위한 HMM WORD SPOTTING 기법을 검토하였다. 실험에 사용한 HMM WORD SPOTTING 기법은 O(n)DP 기법와 OPDP 법이다. 인식시스템은 파라메터로 멜켑스트럼 만을 사용한 경우와 동적 파라메터인 희귀계수를 결합한 경우의 2종류이며, 인식 알고리즘은 O(n)DP 법과 유한상태 오토마타에 의해 구문제어를 실？나 ONE PASS DP 법으로 나눌 수 있다. 또한 인식 단위는 음절과 단어가 혼합된 형태이고 학습은 모두 음절단위로 실시하였으며 연속음성 25문장에 대하여 O(n)DP법과 OPDP법의 인식결과를 비교하여 연속음성 인식에 구문제어 효과를 검증하였다. 실험 결과 평균 인식률이 O(n)DP 의 경우 각각 90.6%, 90.9%, OPDP 의 경우 각각 98.4%, 98.6%로 유한 상태 오토마타에 의한 구문제어를 이용한 평균 7.5%의 인식률이 향상되었다.

• Proceedings of the Acoustical Society of Korea Conference
• /
• 1994.06c
• /
• pp.391-394
• /
• 1994

본 연구에서는 전화선을 통한 음성인식을 위해 저잡음의 실험실 환경에서 수집된 음성 자료를 이용하는 접근을 하였다. 전화 음성과 실험실 음성 간의 특성 차이를 보정하기 위해 선형 회귀 분석법을 이용한 SDCN을 제안하였다. 두 자료간의 보정은 동시 녹음된 실험실 환경의 음성과 전화음성의 SNRDP 따른 두 자료간의 차이를 최소화하는 변환행렬을 구해, 이를 학습자료의 변환에 이용한다. 제안된 방법의 타당성을 확인하기 위해 두가지 인식 알고리즘인 DTW와 이산 HMM 에 대해 실험하였다. DTW를 통한 인식에서개선된 SDCN 에 의한 특징벡터의 변환은 기존의 SDCNDP 따른 특징변환보다 8~17%의 인식률이 향상되었다. 이산 HMM으로 인식할 때는 개선된 SDCNDP 의한 전화음성과 실험실 음성과의 유사도를 보다 잘 나타내기 위해 개선된 SDCN을 적용하고, VQ 코드열 상에서이 코드 사상법을 사용하여 인식률의 향상시켰다.

• Proceedings of the IEEK Conference
• /
• 2000.09a
• /
• pp.141-144
• /
• 2000

일반인들은 음성인식을 이용한 생활보조기구들의 필요성이 적지만 장애인이나 노인들은 가족이나 주변인의 도움을 받지 않고서는 가전제품의 작동이나 전화통화 등과 같은 일을 스스로 하기에는 쉽지 않다. 이러한 사람들에게 각 가정에 널리 보급되어 있는 PC를 이용하여서 타인의 도움을 받지 않고서도 간편하게 사용할 수 있게 음성을 이용한 생활보조기구들 제어에 응용하였다본 음성인식기는 음성의 끝점 검출, 음성의 특징계수 추출, 백터 양자화 학습 및 인식, HMM학습 그리고 HMM인식으로 나누어져 있다. 그리고 그 인식 결과에 따라 생활보조기구등을 제어하였다. 이러한 음성인식기를 만드는 것은 노인이나 장애인들에게 자신이 혼자할수 없는 생활의 편리함을가져다 주기 위함이고 일반정상인에게도 많은 편리함을 가져다 주기 위함이다. 그러나 언어 학습과정에서 노인이나 환자는 학습에 어려움이 있어 적은 학습으로도 인식되어야하는 과제가 남아있다.

• Journal of the Korean Institute of Telematics and Electronics B
• /
• v.33B no.4
• /
• pp.131-140
• /
• 1996

In this paper, we propose the method that extracts the speech feature using the hearing model through signal precessing techniques. The proposed method includes following procedure ; normalization of the short-time speech block by its maximum value, multi-resolution analysis using the discrete wavelet transformation and re-synthesize using thediscrete inverse wavelet transformation, differentiation after analysis and synthesis, full wave rectification and integration. In order to verify the performance of the proposed speech feature in the speech recognition task, korean digita recognition experiments were carried out using both the dTW and the VQ-HMM. The results showed that, in case of using dTW, the recognition rates were 99.79% and 90.33% for speaker-dependent and speaker-independent task respectively and, in case of using VQ-HMM, the rate were 96.5% and 81.5% respectively. And it indicates that the proposed speech feature has the potentials to use as a simple and efficient feature for recognition task.

• Speech Sciences
• /
• v.14 no.1
• /
• pp.151-162
• /
• 2007

This study focuses on the classification of pathological voice using GMM (Gaussian Mixture Model) and compares the results to the previous work which was done by ANN (Artificial Neural Network). Speech data from normal people and patients were collected, then diagnosed and classified into two different categories. Six characteristic parameters (Jitter, Shimmer, NHR, SPI, APQ and RAP) were chosen. Then the classification method based on the artificial neural network and Gaussian mixture method was employed to discriminate the data into normal and pathological speech. The GMM method attained 98.4% average correct classification rate with training data and 95.2% average correct classification rate with test data. The different mixture number (3 to 15) of GMM was used in order to obtain an optimal condition for classification. We also compared the average classification rate based on GMM, ANN and HMM. The proper number of mixtures on Gaussian model needs to be investigated in our future work.

• Proceedings of the Korean Information Science Society Conference
• /
• 2003.04c
• /
• pp.178-180
• /
• 2003

EEG-based brain computer interface(BCI) provides a new communication channel between human brain and computer. EEG data is a multivariate time series so that hidden Markov model (HMM) might be a good choice for classification. However EEG is very noisy data and contains artifacts, so useful features mr expected to improve the performance of HMM. In this paper we addresses the usefulness of principal component features with Hidden Markov model (HHM). We show that some selected principal component features can suppress small noises and artifacts, hence improves classification performance. Experimental study for the classification of EEG data during imagination of a left, right up or down hand movement confirms the validity of our proposed method.

• The Journal of the Acoustical Society of Korea
• /
• v.19 no.2
• /
• pp.83-89
• /
• 2000

In this paper, we proposed a discriminative adaptation method for PMC algorithm and achieved improved speech recognition rate. For the adaptation, we adopted modified PMC(MPMC) which is a variant of PMC and discriminatively adapted the association factor for each mixture of the HMM in the MPMC. From the recognition experiments, the proposed method showed better recognition rate than the conventional PMC. Also, compared with STAR algorithm which is another model parameter compensation method, the proposed method showed superior performance when the SNR is very low and the adaptation data is not sufficient.