• Proceedings of the KSPS conference
• /
• 2007.05a
• /
• pp.127-130
• /
• 2007

The conventional feature recombination technique is very effective in the band-limited noise condition, but in broad-band noise condition, the conventional feature recombination technique does not produce notable performance improvement compared with the full-band system. To cope with this drawback, we introduce a new technique of sub-band likelihood computation in the feature recombination, and propose a new feature recombination method by using this sub-band likelihood computation. Furthermore, the reliable sub-band selection based on the signal-to-noise ratio is used to improve the performance of this proposed feature recombination. Experimental results shows that the average error reduction rate in various noise condition is more than 27% compared with the conventional full-band speaker identification system.

• ETRI Journal
• /
• v.26 no.3
• /
• pp.273-276
• /
• 2004

We propose a novel feature processing technique which can provide a cepstral liftering effect in the log-spectral domain. Cepstral liftering aims at the equalization of variance of cepstral coefficients for the distance-based speech recognizer, and as a result, provides the robustness for additive noise and speaker variability. However, in the popular hidden Markov model based framework, cepstral liftering has no effect in recognition performance. We derive a filtering method in log-spectral domain corresponding to the cepstral liftering. The proposed method performs a high-pass filtering based on the decorrelation of filter-bank energies. We show that in noisy speech recognition, the proposed method reduces the error rate by 52.7% to conventional feature.

• MALSORI
• /
• v.55
• /
• pp.131-142
• /
• 2005

A mismatch between the training and the test conditions often causes a drastic decrease in the performance of the speech recognition systems. In this paper, non-linear transformation techniques based on histogram equalization in the acoustic feature space are studied for reducing the mismatched condition. The purpose of histogram equalization(HEQ) is to convert the probability distribution of test speech into the probability distribution of training speech. While conventional histogram equalization methods consider only the probability distribution of a test speech, for noise-corrupted test speech, its probability distribution is also distorted. The transformation function obtained by this distorted probability distribution maybe bring about miss-transformation of feature vectors, and this causes the performance of histogram equalization to decrease. Therefore, this paper proposes a new method of calculating noise-removed probability distribution by using assumption that the CDF of noisy speech feature vectors consists of component of speech feature vectors and component of noise feature vectors, and this compensated probability distribution is used in HEQ process. In the AURORA-2 framework, the proposed method reduced the error rate by over $44\%$ in clean training condition compared to the baseline system. For multi training condition, the proposed methods are also better than the baseline system.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• v.9 no.2
• /
• pp.473-476
• /
• 2005

This paper is proposed a robust various feature parameters in noise. Feature parameter MFCC(Mel Frequency Cepstral Coefficient) used in conventional speech recognition shows good performance. But, parameter transformed feature space that uses PCA(Principal Component Analysis)and ICA(Independent Component Analysis) that is algorithm transformed parameter MFCC's feature space that use in old for more robust performance in noise is compared with the conventional parameter MFCC's performance. The result shows more superior performance than parameter and MFCC that feature parameter transformed by the result ICA is transformed by PCA.

• ETRI Journal
• /
• v.46 no.1
• /
• pp.22-34
• /
• 2024

Exposure to varied noisy environments impairs the recognition performance of artificial intelligence-based speech recognition technologies. Degraded-performance services can be utilized as limited systems that assure good performance in certain environments, but impair the general quality of speech recognition services. This study introduces an audiovisual speech recognition (AVSR) model robust to various noise settings, mimicking human dialogue recognition elements. The model converts word embeddings and log-Mel spectrograms into feature vectors for audio recognition. A dense spatial-temporal convolutional neural network model extracts features from log-Mel spectrograms, transformed for visual-based recognition. This approach exhibits improved aural and visual recognition capabilities. We assess the signal-to-noise ratio in nine synthesized noise environments, with the proposed model exhibiting lower average error rates. The error rate for the AVSR model using a three-feature multi-fusion method is 1.711%, compared to the general 3.939% rate. This model is applicable in noise-affected environments owing to its enhanced stability and recognition rate.

• The Journal of the Acoustical Society of Korea
• /
• v.25 no.2E
• /
• pp.69-73
• /
• 2006

In this paper, we propose a robust music audio fingerprinting system for automatic music retrieval. The fingerprint feature is extracted from the long-term dynamic modulation spectrum (LDMS) estimation in the perceptual compressed domain. The major advantage of this feature is its significant robustness against severe background noise from the street and cars. Further the fast searching is performed by looking up hash table with 32-bit hash values. The hash value bits are quantized from the logarithmic scale modulation frequency coefficients. Experiments illustrate that the LDMS fingerprint has advantages of high scalability, robustness and small fingerprint size. Moreover, the performance is improved remarkably under the severe recording-noise conditions compared with other power spectrum-based robust fingerprints.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.45 no.6
• /
• pp.148-155
• /
• 2008

In this paper, we constructed the noise robust content-based music retrieval system in mobile environment. The performance of the proposed system was verified with ZCPA feature which is blown to have noise robust characteristic in speech recognition application. In addition, new indexing and fast retrieval method are proposed to improve retrieval speed about 99% compare to exhaustive retrieval for large music DB. From the computer simulation results in noise environment of 15dB - 0dB SNR, we confirm the superior performance of the proposed system about 5% - 30% compared to MFCC and FBE(filter bank energy) feature.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.8 no.2
• /
• pp.359-364
• /
• 2004

In this paper, we implemented real-time speaker undependent speech recognizer that is robust in noise environment using DSP(Digital Signal Processor). Implemented system is composed of TMS320C32 that is floating-point DSP of Texas Instrument Inc. and CODEC for real-time speech input. Speech feature parameter of the speech recognizer used robust feature parameter in noise environment that is transformed feature space of MFCC(met frequency cepstral coefficient) using ICA(Independent Component Analysis) on behalf of MFCC. In recognition result in noise environment, we hew that recognition performance of ICA feature parameter is superior than that of MFCC.

• The Journal of the Acoustical Society of Korea
• /
• v.18 no.6
• /
• pp.57-64
• /
• 1999

In this paper, we investigated the robust endpoint detection algorithm in noisy environment. A new feature parameter based on a discrete wavelet transform is proposed for word boundary detection of isolated utterances. The sum of standard deviation of wavelet coefficients in the third coarse and weighted first detailed scale is defined as a new feature parameter for endpoint detection. We then developed a new and robust endpoint detection algorithm using the feature found in the wavelet domain. For the performance evaluation, we evaluated the detection accuracy and the average recognition error rate due to endpoint detection in an HMM-based recognition system across several signal-to-noise ratios and noise conditions.

• Phonetics and Speech Sciences
• /
• v.2 no.1
• /
• pp.77-85
• /
• 2010

This paper proposes an efficient feature vector processing technique to guard the Speech Emotion Recognition (SER) system against a variety of noises. In the proposed approach, emotional feature vectors are extracted from speech processed by comb filtering. Then, these extracts are used in a robust model construction based on feature vector classification. We modify conventional comb filtering by using speech presence probability to minimize drawbacks due to incorrect pitch estimation under background noise conditions. The modified comb filtering can correctly enhance the harmonics, which is an important factor used in SER. Feature vector classification technique categorizes feature vectors into either discriminative vectors or non-discriminative vectors based on a log-likelihood criterion. This method can successfully select the discriminative vectors while preserving correct emotional characteristics. Thus, robust emotion models can be constructed by only using such discriminative vectors. On SER experiment using an emotional speech corpus contaminated by various noises, our approach exhibited superior performance to the baseline system.