China
Africa
Explore chapters and articles related to this topic
Digital Techniques for Mobile Radio
Published in Michel Daoud Yacoub, Foundations of Mobile Radio Engineering, 2019
Linear prediction is a form of estimation using a linear combination of present and past samples of a stationary process to predict a sample of the process in the future. Let Sn-k, k = 1 be random samples from a stationary process S(t) and let Sn be the sample to be predicted. The estimate of Sn is Sn such that () S^n=∑k=1MhkSn−k
Basic principles
Published in Michael Talbot-Smith, Audio Engineer's Reference Book, 2013
John Ratcliff, Talbot-Smith Michael, J. Patrick Wilson, Louis D. Fielder, Glynne Parry, Richard Tyler, Michael Gayford, Roger Derry
Linear prediction Linear prediction is a technique that uses previous time samples to predict the value of a new sample. This has the advantage of allowing only a reduced amplitude error signal to be encoded rather than the original signal and minimizes the error generated by the quantization process. A better understanding of linear prediction is obtained by examination of what happens in the spectral domain. Prediction can be looked at as a filtering process that has a frequency-response characteristic that is the inverse of the spectral characteristic of the input signal. When the input signal is treated in this manner its overall level is reduced, its spectrum is flattened and its coding errors are reduced. This process is often called removing the redundancy from the signal because any part of the signal that can be predicted does not need to be transmitted from encoder to decoder, and therefore is redundant.
Speech Signal Processing
Published in Richard C. Dorf, Circuits, Signals, and Speech and Image Processing, 2018
Jerry D. Gibson, Bo Wei, Hui Dong, Yariv Ephraim, Israel Cohen, Jesse W. Fussell, Lynn D. Wilcox, Marcia A. Bush
where s(n) is the output and u(n) is the input (perhaps unknown). The model parameters are a(k) for k = 1, p, b(l) for l = 1, q, and G. b(0) is assumed to be unity. This model, described as an autoregressive moving average (ARMA) or pole-zero model, forms the foundation for the analysis method termed linear prediction. An autoregressive (AR) or all-pole model, for which all of the “b” coefficients except b(0) are zero, is frequently used for speech analysis (Markel and Gray, 1976).
A comparative study on estimation of fractal dimension of EMG signal using SWT and FLP
Published in Computer Methods in Biomechanics and Biomedical Engineering: Imaging & Visualization, 2023
A. A. Navish, M. Priya, R. Uthayakumar
This procedure predicts the transmission coefficient of the predictor and we can model the waveform without any artifact activities. Accuracy of signal waves can be represented by linear prediction method (Ortigueira et al. 2002). We admire the FLP technique to show that so many habitual life wave forms and several phenomena have immanently fractional order dynamics and the ways which are based on fractional order calculus techniques are more appropriate to model these signals with a higher accuracy value. The field of digital signal processing has a significant utilisation of a concept called linear prediction. Even though this method commenced in the middle of the 20th century, the application of this particular technique can be seen today in advancements with a generalisation termed FLP. Mainly, this theory gains attention in computer generation of waveforms, representation, modelling and compression.
Acoustic–Phonetic Analysis for Speech Recognition: A Review
Published in IETE Technical Review, 2018
Biswajit Dev Sarma, S. R. Mahadeva Prasanna
There are mainly three types of approaches in the literature for glottal activity detection. They are time-domain, frequency-domain, and statistical approaches. In time-domain and frequency-domain approaches, acoustic features representing the production characteristics of voiced speech were measured. Parameters such as short-term zero crossing rate, the first linear prediction (LP) coefficient, autocorrelation coefficient at the first lag, long-term normalized autocorrelation peak strength, normalized LP error, harmonic measure from the instantaneous frequency amplitude spectrum, cepstral peak strength, normalized low-frequency energy were used to measure some production characteristics related to energy, periodicity and short-term correlation [11–13]. Most of these methods rely on threshold setting which are critical in detecting glottal activity.
Damage detection of a cable-stayed bridge using feature extraction and selection methods
Published in Structure and Infrastructure Engineering, 2019
Hossein Babajanian Bisheh, Gholamreza Ghodrati Amiri, Masoud Nekooei, Ehsan Darvishan
All the features are computed from short-time windows with an overlap between adjacent windows extracted from the vibration acceleration signals. These statistical indicators include standard deviation (T1), root mean square (RMS) (T2), productivity ratio (the measure of how well the signal can be predicted by order of linear prediction) (T3), autocorrelation coefficient (T4), and zero crossing rate (the rate of sign-changes along a signal, that is, the rate at which the signal changes from positive to negative or back) (T5). These parameters are calculated using Equations (1) and (2) as described in Table 1.