Zero-crossing based spectral analysis and SVD spectral analysis for formant frequency estimation in noise

Sreenivas, T.V.; Niederjohn, Russell J.

doi:10.1109/78.124939

Cited by 59 publications

(25 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The expression also indicates that the higher level-crossing instants at lower frequencies have larger variance compared to lower level-crossing instants at higher frequencies. The above analysis is a generalization of the results in [12] and [13] to time-varying signals. For the simpler case of a constant-frequency sinusoid, is least in the case of ZC instants and it increases with increasing level value.…”

Section: A Perturbation Analysismentioning

confidence: 66%

“…Let the corresponding level-crossing instant for the clean signal be with denoting the perturbation in the LC instant. Thus, we have sin (9) sin (10) We model the perturbation as a zero-mean, uncorrelated random variable [13]. Consider (11) Using a second-order Taylor series approximation to evaluate the expectations involved, we obtain, after simplification (assuming stationary additive noise, high SNR, and )…”

Section: A Perturbation Analysismentioning

confidence: 99%

“…The bandpass filter serves to cut off Consider the following specification for the TV bandpass filter (the notation means that is discrete and is continuous): for otherwise (13) where is the center frequency of the filter at time , and is the spread/bandwidth about . The output of such a system to an input, is given by (14) where is the window function used in obtaining the STFT of denoted by and defined as .…”

Section: Performance Improvementmentioning

confidence: 99%

“…This is not designed to satisfy the general properties of a TFD, but aimed at mimicking the biological functions while processing complex stimuli using signal processing techniques. A modification of this approach uses only zero-crossing information [12], [13]. These techniques assume that the signals are quasistationary (spectrum slowly changing with time) and, hence, can be analyzed, similar to short-time Fourier transform (STFT) on a frame-by-frame basis.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Auditory motivated level-crossing approach to instantaneous frequency estimation

Sekhar

Sreenivas

2005

IEEE Trans. Signal Process.

Self Cite

View full text Add to dashboard Cite

Abstract-We address the problem of estimating the instantaneous frequency (IF) of a phase signal using its level-crossing (LC) information based on front-end auditory processing motivation. We show that the problem of IF estimation using LC information can be cast in the framework of estimation from irregularly sampled data. The formulation has the generality of estimating different types of IF without the need for a quasistationary assumption. We consider two types of IF-polynomial and bandlimited; we use polynomial interpolating functions for the former, and for the latter, we propose a novel "line plus sum of sines" model.

show abstract

Section: A Perturbation Analysismentioning

confidence: 66%

Section: A Perturbation Analysismentioning

confidence: 99%

Section: Performance Improvementmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Auditory motivated level-crossing approach to instantaneous frequency estimation

Sekhar

Sreenivas

2005

IEEE Trans. Signal Process.

Self Cite

View full text Add to dashboard Cite

show abstract

“…The corresponding algorithm (cf. Fig 2) can be compared to those used in auditory feature extraction, like ZCPA that is used for speech recognition (Ghitza, 1994;Kim et al, 1999;Sreenivas and Niederjohn, 1992). It provides a quantity homogeneous to the ENL which we call the "Instantaneous Mean Power" or IMP feature.…”

Section: Feature Extractionmentioning

confidence: 99%

Tactile Texture Discrimination in the Robot-Rat Psikharpax

Pirim

Meyer

2010

Proceedings of the Third International Conference on Bio-Inspired Systems and Signal Processing

View full text Add to dashboard Cite

Abstract:We endowed a whiskered robot with a simple algorithm allowing to discriminate textures. Its efficiency and robustness have been demonstrated using both a fixed head and a mobile platform. Comparatively to previous similar approaches, this system affords greater behavioral capacities and proves to be able to complement or supply vision in simple navigation tasks. The corresponding results suggest that the length and number of the whiskers involved play a role in texture discrimination. They also suggest that two hypotheses that are currently considered as mutually exclusive to explain texture recognition in rats -i.e., the "kinetic signature hypothesis" and the "resonance hypothesis" -may be, in fact, complementary.

show abstract

Improvement of noise robustness for formant frequency extraction based on linear predictive analysis

Zhao

Shimamura

Takahashi³

et al. 2002

Electron Comm Jpn Pt III

View full text Add to dashboard Cite

SUMMARYIn this paper, we study improving the noise robustness of formant frequency extraction based on linear predictive analysis by using the noise reduction ability of the autocorrelation function. The autocorrelation function has the property of concentrating the noise components in the short delayed parts of speech signals corrupted by white noise. We believe that the noise robustness of formant frequency extraction can be improved by utilizing the above property and using the autocorrelation function of the speech signal instead of the signal itself in formant frequency extraction. In this paper, we analyze the principle behind extracting the formant frequency from the autocorrelation function and examine this problem and a possible solution in detail. Then based on this analysis, we propose Method 1 that performs linear predictive analysis of the autocorrelation function of the speech signal. Then we verify from the experimental results that an extraction accuracy at the same level as the conventional method is obtained for a clean signal by Method 1, and the extraction accuracy is significantly improved over the conventional method for speech corrupted by noise. However, inadequate extraction accuracy is also indicated in a very noisy environment, and we analyze the cause and propose Method 2 as an improved method that subtracts the autocorrelation function. The experimental results show that when the signal-to-noise ratio is 15 dB or less, the formant frequency extraction error (Average Absolute Error) in Method 2 is kept to about one-third the error of the conventional method and about one-half the error of Method 1.

show abstract

Zero-crossing based spectral analysis and SVD spectral analysis for formant frequency estimation in noise

Cited by 59 publications

References 21 publications

Auditory motivated level-crossing approach to instantaneous frequency estimation

Auditory motivated level-crossing approach to instantaneous frequency estimation

Tactile Texture Discrimination in the Robot-Rat Psikharpax

Improvement of noise robustness for formant frequency extraction based on linear predictive analysis

Contact Info

Product

Resources

About