Cepstrum

About: Cepstrum is a research topic. Over the lifetime, 3346 publications have been published within this topic receiving 55742 citations.

Evaluation of the SPLICE algorithm on the Aurora2 database.

Abstract: This paper describes recent improvements to SPLICE, Stereobased Piecewise Linear Compensation for Environments, which produces an estimate of cepstrum of undistorted speech given the observed cepstrum of distorted speech For distributed speech recognition applications, SPLICE can be placed at the server, thus limiting the processing that would take place at the client We evaluated this algorithm on the Aurora2 task, which consists of digit sequences within the TIDigits database that have been digitally corrupted by passing them through a linear filter and/or by adding different types of realistic noises at SNRs ranging from 20dB to -5dB On set A data, for which matched training data is available, we achieved a 66% decrease in word error rate over the baseline system with clean models This preliminary result is of practical significance because in a server implementation, new noise conditions can be added as they are identified once the service is running

The modified group delay function and its application to phoneme recognition

Abstract: We explore a new spectral representation of speech signals through group delay functions. The group delay functions by themselves are noisy and difficult to interpret owing to zeroes that are close to the unit circle in the z-domain and these clutter the spectra. A new modified group delay function (Yegnanarayan, B. and Murthy, H.A., IEEE Trans. Sig. Processing, vol.40, p.2281-9, 1992) that reduces the effects of zeroes close to the unit circle is used. Assuming that this new function is minimum phase, the modified group delay spectrum is converted to a sequence of cepstral coefficients. A preliminary phoneme recogniser is built using features derived from these cepstra. Results are compared with those obtained from features derived from the traditional mel frequency cepstral coefficients (MFCC). The baseline MFCC performance is 34.7%, while that of the best modified group delay cepstrum is 39.2%. The performance of the composite MFCC feature, which includes the derivatives and double derivatives, is 60.7%, while that of the composite modified group delay feature is 57.3%. When these two composite features are combined, /spl sim/2% improvement in performance is achieved (62.8%). When this new system is combined with linear frequency cepstra (LFC) (Gadde, V.R.R. et al., The SRI SPINE 2001 Evaluation System. http://elazar.itd.nrl.navy.mil/spine/sri2/presentation/sri2001.html, 2001), the system performance results in another /spl sim/0.8% improvement (63.6%).

Restoration of medical ultrasound images using two-dimensional homomorphic deconvolution

Abstract: Describes how two-dimensional (2D) homomorphic deconvolution can be used to improve the lateral and radial resolution of medical ultrasound images recorded by a sector scanner. The recorded radio frequency ultrasound image in polar coordinates is considered as a 2D sequence of angle and depth convolved with a 2D space invariant point-spread function (PSF). Each polar coordinate sequence is transformed into the 2D complex cepstrum domain using the fast Fourier transform for Cartesian coordinates. The low-angle and low-depth portion of this sequence is taken as an estimate of the complex cepstrum representation of the PSF. It is transformed back to the Fourier frequency domain and is used to compute the deconvolved angle and depth sequence by 2D Wiener filtering. Two-dimensional homomorphic deconvolution produced substantial improvement in the resolution of B-mode images of a tissue-mimicking phantom in vitro and of several human tissues in vivo. It was better than lateral or radial homomorphic deconvolution alone, and better than 2D Wiener filtering with a PSF recorded in vitro. >

Application of autoregressive spectral analysis to cepstral estimation of mean scatterer spacing

Abstract: The problem of estimation of mean scatterer spacing in an object containing regularly spaced structures is addressed. An autoregressive (AR) spectral estimation method is compared with a conventional fast Fourier transform (FFT)-based approach for this task. Regularly spaced structures produce a periodicity in the power spectrum of ultrasonic backscatter. This periodicity is manifested as a peak in the cepstrum. A phantom was constructed for comparison of the two methods. It contained regularly spaced nylon filaments. It also contained randomly positioned glass spheres that produced incoherent backscatter. In an experiment in which this target was interrogated using broadband ultrasound, the AR spectral estimate offered considerable improvement over the FFT when the analysis gate length was on the order of the structural dimension. Advantages included improved resolution, reduction in bias and variance of scatterer spacing estimates, and greater resistance to ringing artifacts. Data were also acquired from human liver in vivo. AR spectral estimates on human data exhibited a decreased dependence on gate length. These results offer promise for enhanced spatial resolution and accuracy in ultrasonic tissue characterization and nondestructive evaluation of materials. >