Evaluation of Quantile Based Histogram Equalization in Combination with Different Root Functions
نویسندگان
چکیده
This paper presents an evaluation of the RWTH large vocabulary speech recognition system on the Aurora 4 noisy Wall Street Journal database. First, the influence of different root functions replacing the logarithm in the feature extraction is studied. Then quantile based histogram equalization is applied, a parametric method to increase the noise robustness by reducing the mismatch between the training and test data distributions. Putting everything together, the word error rate could be reduced from 45.7% to 25.5% (clean training data) and from 19.5% to 17.0% (multicondition training data). Logarithm and Root Functions In a conventinal Mel-frequency cepstral coefficient (MFCC) feature extraction a logarithm is applied after the Mel-scaled filterbank to reduce the dynamic range of the signal. This logarithm can be replaced by a root function. The general relation between root/power functions and the logarithm can be expressed as follows:
منابع مشابه
Density-Based Histogram Partitioning and Local Equalization for Contrast Enhancement of Images
Histogram Equalization technique is one of the basic methods in image contrast enhancement. Using this method, in the case of images with uniform gray levels (with narrow histogram), causes loss of image detail and the natural look of the image. To overcome this problem and to have a better image contrast enhancement, a new two-step method was proposed. In the first step, the image histogram is...
متن کاملEvaluation of quantile based histogram equalization with filter combination on the Aurora 3 and 4 databases
The recognition performance of automatic speech recognition systems can be improved by reducing the mismatch between training and test data during feature extraction. The approach described in this paper is based on estimating the signal’s cumulative density functions on the filter bank using a small number of quantiles. A two–step transformation is then applied to reduce the difference between...
متن کاملHigh Speed Quantile Based Histogram Equalization
8 In this paper we introduce a new histogram equalization based contrast enhancement method called High Speed Quantile Based Histogram Equalization (HSQHE) suitable for high contrast digital images. The proposed method is an effective tool to deal with the “mean-shift” problem, which is a usual problem with the histogram equalization based contrast enhancement methods. The main idea of HSQHE is...
متن کاملCombining neighboring filter channels to improve quantile based histogram equalization
A mismatch between the training data and the test condition of an automatic speech recognition system usually deteriorates the recognition performance. Quantile based histogram equalization can increase the system’s robustness by approximating the cumulative density function of the current signal and then reducing an eventual mismatch based on this estimate. In a first step each output of the m...
متن کاملHSI based colour image equalization using iterative nth root and nth power
In this paper an equalization technique for colour images is introduced. The method is based on n th root and n th power equalization approach but with optimization of the mean of the image in different colour channels such as RGB and HSI. The performance of the proposed method has been measured by the means of peak signal to noise ratio. The proposed algorithm has been compared with convention...
متن کامل