• Speech Sciences
• /
• v.10 no.4
• /
• pp.107-115
• /
• 2003

In this paper we tried to classify the pathological voice signal with severe noise component based on two different parameters, the spectral slope and the ratio of energies in the harmonic and noise components (HNR), The spectral slope is obtained by using a curve fitting method and the HNR is computed in cepstrum quefrency domain. Speech data from normal peoples and patients are collected, diagnosed and divided into three different classes (normal, relatively less noisy and severely noisy data), The mean values and the standard deviations of the spectral slope and the HNR are computed and compared with in the three kinds of data to characterize and classify the severely noisy pathological voice signals from others.

• Proceedings of the KIEE Conference
• /
• 1987.07b
• /
• pp.1583-1586
• /
• 1987

In the study for effective speech control we designed a personal computer system with A/D converter in which the speech signal is transformed by digital data displayed graphically on the moniter and with a D/A converter in which the digital data is transformed into speech signal which people can hear. We analyzed the character of the speech signal produced by the system. We designed the adaptive noise cancel algorithm so that noise and Interference are cancelled whenever the speech signal is recognized by the computer system. This is a basic system for artificial Intelligence.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.15 no.8
• /
• pp.1665-1674
• /
• 2011

In this paper, we proposed an algorithm for speech enhancement of speeches corrupted by colored noise. When there is no correlation between colored noise and speech signal, the colored noise turns into white noise through whitening transformation. This transformed signal has been applied to the generalized subspace approach for speech enhancement. The speech spectral distortion, produced by the whitening transformation as pre-processing, has been restored by using the inverse whitening transformation as post-processing of the proposed algorithm. The performance of the proposed algorithm for speech enhancement has been confirmed by computer simulation. The colored noises used in this experiment were car noise and multi-talker babble. It is confirmed that the proposed algorithm shows better performance from SNR and SSD viewpoint over the previous approach with the data from the AURORA and TIMIT data base.

• Journal of Information Processing Systems
• /
• v.15 no.4
• /
• pp.748-764
• /
• 2019

In order to deal with the filtering delay problem of least mean square adaptive filter noise reduction algorithm and music noise problem of spectral subtraction algorithm during the speech signal processing, we combine these two algorithms and propose one novel noise reduction method, showing a strong performance on par or even better than state of the art methods. We first use the least mean square algorithm to reduce the average intensity of noise, and then add spectral subtraction algorithm to reduce remaining noise again. Experiments prove that using the spectral subtraction again after the least mean square adaptive filter algorithm overcomes shortcomings which come from the former two algorithms. Also the novel method increases the signal-to-noise ratio of original speech data and improves the final noise reduction performance.

• Speech Sciences
• /
• v.10 no.3
• /
• pp.263-277
• /
• 2003

One of major problems in speech recognition is performance degradation due to the mismatch between the training and test environments. Recently, Stereo-based Piecewise LInear Compensation for Environments (SPLICE), which is frame-based bias removal algorithm for cepstral enhancement using stereo training data and noisy speech model as a mixture of Gaussians, was proposed and showed good performance in noisy environments. In this paper, we propose several methods to improve the conventional SPLICE. First we apply Cepstral Mean Subtraction (CMS) as a preprocessor to SPLICE, instead of applying it as a postprocessor. Secondly, to compensate residual distortion after SPLICE processing, two-stage SPLICE is proposed. Thirdly we employ phonetic information for training SPLICE model. According to experiments on the Aurora 2 database, proposed method outperformed the conventional SPLICE and we achieved a 50% decrease in word error rate over the Aurora baseline system.

• Journal of the Korea Society of Computer and Information
• /
• v.29 no.1
• /
• pp.41-49
• /
• 2024

Speech recognition technology is continuously advancing and widely used in various fields. In this study, we aimed to investigate the impact of speech data quality on speech recognition models by dividing the dataset into the entire dataset and the top 70% based on Signal-to-Noise Ratio (SNR). Utilizing Seamless M4T and Google Cloud Speech-to-Text, we examined the text transformation results for each model and evaluated them using the Levenshtein Distance. Experimental results revealed that Seamless M4T scored 13.6 in models using data with high SNR, which is lower than the score of 16.6 for the entire dataset. However, Google Cloud Speech-to-Text scored 8.3 on the entire dataset, indicating lower performance than data with high SNR. This suggests that using data with high SNR during the training of a new speech recognition model can have an impact, and Levenshtein Distance can serve as a metric for evaluating speech recognition models.

• Proceedings of the KIEE Conference
• /
• 2000.07d
• /
• pp.3166-3168
• /
• 2000

This paper proposed variable threshold dual rate ADPCM coding method which is modified from the standard ADPCM of ITU G.726 for speech quality improvement. The speech quality of variable threshold dual rate ADPCM is better than single rate ADPCM at noisy environment without increasing the complexity by using ZCR(Zero Crossing Rate). In this case, ZCR is used to divide input signal samples into two categories(noisy & speech). The samples with higher ZCR is categorized as the noisy region and the samples with lower ZCR is categorized as the speech region. Noisy region uses higher threshold value to be compressed by 16Kbps for reduced bit rates and the speech region uses lower threshold value to be compressed by 40Kbps for improved speech quality. Comparing with the conventional ADPCM, which adapts the fixed coding rate. the proposed variable threshold dual rate ADPCM coding method improves noise character without increasing the bit rate. For real time applications, ZCR calculation was considered as a simple method to obtain the background noise information for preprocess of speech analysis such as FFT and the experiment showed that the simple calculation of ZCR can be used without complexity increase. Dual rate ADPCM can decrease the amount of transferred data efficiently without increasing complexity nor reducing speech quality. Therefore result of this paper can be applied for real-time speech application such as the internet phone or VoIP.

• IEIE Transactions on Smart Processing and Computing
• /
• v.1 no.3
• /
• pp.143-151
• /
• 2012

A new data-driven method for the design of a blind modulation frequency filter that suppresses the slow-varying noise components is proposed. The proposed method is based on the temporal local decorrelation of the feature vector sequence, and is done on an utterance-by-utterance basis. Although the conventional modulation frequency filtering approaches the same form regardless of the task and environment conditions, the proposed method can provide an adaptive modulation frequency filter that outperforms conventional methods for each utterance. In addition, the method ultimately performs channel normalization in a feature domain with applications to log-spectral parameters. The performance was evaluated by speaker-independent isolated-word recognition experiments under additive noise environments. The proposed method achieved outstanding improvement for speech recognition in environments with significant noise and was also effective in a range of feature representations.

• 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.

• ETRI Journal
• /
• v.39 no.1
• /
• pp.21-29
• /
• 2017

This paper addresses the problem of unsupervised speech separation based on robust non-negative matrix factorization (RNMF) with ${\beta}$-divergence, when neither speech nor noise training data is available beforehand. We propose a robust version of non-negative matrix factorization, inspired by the recently developed sparse and low-rank decomposition, in which the data matrix is decomposed into the sum of a low-rank matrix and a sparse matrix. Efficient multiplicative update rules to minimize the ${\beta}$-divergence-based cost function are derived. A convolutional extension of the proposed algorithm is also proposed, which considers the time dependency of the non-negative noise bases. Experimental speech separation results show that the proposed convolutional RNMF successfully separates the repeating time-varying spectral structures from the magnitude spectrum of the mixture, and does so without any prior training.