Academic literature on the topic 'Cepstrum'

A method of cepstrum analysis is developed for the purpose of resolving thin-beds. The method relies on the detection of periodic pulses of the cepstra of reflectivity functions, which are isolated by computing a sub-cepstrum and a sum-cepstrum, and highlighted with a discriminator, where the sub-cepstrum of the functions f₁(t) and f₂(t) is the difference between the cepstra of the two functions, the sum-cepstrum of f₁(t) is the sum of the sub-cepstra of f₁(t) and f_k(t), k=2,3,4,... , and the discriminator is the product of the sum-cepstrum and the autocovariance of the sum-cepstrum. The technique requires at least two reflected wavelets generated by the same source. The method was applied to synthetic thin lens models. The method is shown to be sensitive to the ratio of the reflection coefficients at the top and bottom of the thin-bed. Specifically, the resolution depends on the ratio of the reflection coefficients. Optimum resolution is achieved when the reflection coefficients at the top and bottom of the thin-bed are equal in absolute magnitude. In addition, in the noise-free case, the absolute magnitude of the cepstral pulses can be used to determine the absolute magnitude of the ratio of the reflection coefficients. The technique is also sensitive to the sample interval used. The finest sample interval provides the best resolution because it produces the sharpest cepstral pulses and resolves the thinnest beds. The resolution of the method is drastically reduced by random noise, although thin-bed thicknesses are still detectable when the S/N of the synthetic seismic section is 15/1 and the upper frequency of the bandwidth of the noise is 1.1 octaves above the upper frequency of the bandwidth of the source wavelet.<br>Master of Science

Thesis (M.S. in Electrical Engineering) Naval Postgraduate School, March 1995.<br>"March 1995." Thesis advisor(s): Monique Fargue, Ralph Hippenstiel. Includes bibliographical references. Also available online.

This work describes development of a diff tool working with audio files containing general sound such as music, speech and other sounds. There are presented facts from different domains of science related to sound, such as psychoacoustics, speech recognition and automatic music genre categorisation. This paper also contains description of some diff algorithms and external tools needed for development of the goal application. Moreover, there is introduced design and implementation of the application, settings used for sound features extraction and evaluation of attained results.

Solving speech recognition problems requires an adequate feature extraction technique to transform the raw speech signal to a set of feature vectors to preserve most of information corresponding to the speech signal. The features should ideally be compact, distinct and well representative of the speech signal. If the feature vectors do not represent the important content of the speech, the performance of the system will perform poorly regardless of the pattern recognition techniques applied. Many different feature extraction representations of the speech signal have been suggested and tried for speech recognition. The most popular features which are used currently are Mel- frequency cesptral coefficients (MFCC) and perceptual linear prediction (PLP), which are based on one dimensional cepstrum analysis. The two dimensional cepstrum (TDC) is an alternative approach for time-frequency representation of any speech signal which can preserve both the instantaneous and transitional information of the speech signal. Here, in this thesis, the principle aim concerns the study of the two dimensional cepstrum analysis as a feature extraction technique for speech recognition. A novel feature extraction technique, two dimensional root cepstrum (TDRC) is also introduced. It has the advantage of an adjustable y parameter which can be used to optimise the feature extraction process, reducing the dimensions of the feature matrix and giving simple computation. In addition, the Mel TDRC has been proposed as a modified method of original TDRC to improve the accuracy. It is shown that both the TDC and the TDRC outperform the conventional cepstrum. To preserve both magnitude and phase details of the speech signal simultaneously in a feature matrix, the Hartley transform (HT) is suggested as a substitute for the Fourier transform (FT) in two-dimensional cepstrum analysis. Experimental results demonstrate the enhanced capability of the HT in the two dimensional root cepstral analysis to improve recognition accuracy. An experimental comparative study of 9 kinds of feature extraction methods based on cepstral analysis are also carried out.

<p> Voice recognition systems have found widespread use in applications such as tele-shopping, tele-banking, information services, home automation, voice message security, and voice call dialing, which allows a driver to make calls safely while driving. </p><p> This project presents the development of a high performance speech recognition system using human voice models. Recognizing the behavior of the human ear, the Mel Frequency Cepstral Coefficient (MFCC) method is used to develop the system capability for feature extraction. Vector quantization optimized by the Linde Buzo Gray (LGB) algorithm is used for feature matching. Experimental results show that the system has over 90% success rate in the noise-free case, but the system performance deteriorates in the presence of noise. The system, however, has better recognition ability when the noise signal consists of harmonic components, as compared to a non-stationary, non-harmonic signal.</p>

The diploma thesis Detection of Alcohol in Speech Signal first describes the effect of alcohol on the human body. The second part deals with ways to obtain parameters that describe the speech signal. The third part provides a brief overview of previous case studies and patents focused on the detection of alcohol in the speech signal. The fourth part presents the collected own database of voice recordings and developed software application for the analysis of intoxicated speech. The final part describes the measured changes in speech signal parameters that indicate alcohol intoxication.

Speech as a medium for communication has always been important in its ability to convey our ideas, personality and emotions. It is therefore not strange that Quality of Experience (QoE) becomes central to any business relying on voice communication. Using Unified Communication (UC) systems, users can communicate with each other in several ways using many different devices, making QoE an important aspect for such systems. For this thesis, automatic methods for assessing speech quality of the voice calls in Briteback’s UC application is studied, including a comparison of the researched methods. Three methods all using a Gaussian Mixture Model (GMM) as a regressor, paired with extraction of Human Factor Cepstral Coefficients (HFCC), Gammatone Frequency Cepstral Coefficients (GFCC) and Modified Mel Frequency Cepstrum Coefficients (MMFCC) features respectively is studied. The method based on HFCC feature extraction shows better performance in general compared to the two other methods, but all methods show comparatively low performance compared to literature. This most likely stems from implementation errors, showing the difference between theory and practice in the literature, together with the lack of reference implementations. Further work with practical aspects in mind, such as reference implementations or verification tools can make the field more popular and increase its use in the real world.