Contemporary Methods for Speech Parameterization by Todor Ganchev

By Todor Ganchev

Contemporary equipment for Speech Parameterization bargains a common view of short-time cepstrum-based speech parameterization and offers a standard floor for extra in-depth experiences at the topic. in particular, it bargains a accomplished description, comparative research, and empirical functionality overview of 11 modern speech parameterization equipment, which compute short-time cepstrum-based speech gains.

Among those are 5 discrete wavelet packet rework (DWPT)-based, six discrete Fourier rework (DFT)-based speech beneficial properties and a few in their editions that have been used at the speech reputation, speaker acceptance, and different similar speech processing projects. the most similarities and ameliorations of their computation are mentioned and empirical effects from functionality assessment in universal experimental stipulations are provided. the popularity accuracy received at the monophone reputation, non-stop speech attractiveness and speaker popularity projects is contrasted opposed to the single received for the well known and known Mel Frequency Cepstral Coefficients (MFCC).

It is proven that a lot of those equipment bring about speech beneficial properties that do supply aggressive functionality on a definite speech processing setup compared to the venerable MFCC. The final doesn't objective the advertising of convinced speech gains yet as an alternative goals to reinforce the typical knowing in regards to the merits and downsides of many of the speech parameterization thoughts on hand at the present time and to supply the foundation for number of a suitable speech parameterization in each one specific case.

Show description

Read Online or Download Contemporary Methods for Speech Parameterization PDF

Similar human-computer interaction books

Human-Computer Interaction: An Empirical Research Perspective

Human-Computer interplay: An Empirical learn point of view is the definitive consultant to empirical examine in HCI. The booklet starts off with foundational themes together with historic context, the human issue, interplay parts, and the basics of technology and examine. From there, you'll growth to studying concerning the tools for carrying out an test to guage a brand new laptop interface or interplay strategy.

Understanding Mobile Human-Computer Interaction

Taking a mental point of view, this e-book examines the position of Human-Computer interplay within the box of knowledge structures learn. The introductory component of the e-book covers the fundamental tenets of the HCI self-discipline, together with the way it constructed and an summary of many of the educational disciplines that give a contribution to HCI study.

Introducing Spoken Dialogue Systems into Intelligent Environments

Introducing Spoken discussion platforms into clever Environments outlines the formalisms of a unique knowledge-driven framework for spoken discussion administration and offers the implementation of a model-based Adaptive Spoken discussion Manager(ASDM) known as OwlSpeak. The authors have pointed out 3 stakeholders that in all likelihood effect the habit of the ASDM: the consumer, the SDS, and a posh clever surroundings (IE) which includes quite a few units, prone, and job descriptions.

Emerging Research and Trends in Interactivity and the Human-Computer Interface

With numerous rising and cutting edge applied sciences mixed with the lively participation of the human point because the significant connection among the tip person and the electronic realm, the pervasiveness of human-computer interfaces is at an all time excessive. rising examine and tendencies in Interactivity and the Human-Computer Interface addresses the most problems with curiosity in the tradition and layout of interplay among people and desktops.

Extra info for Contemporary Methods for Speech Parameterization

Example text

2001). In brief, let us denote with n the discrete-time index, and with xðnÞ a discrete-time speech signal that has been sampled with sampling frequency fs . Let us consider that the signal xðnÞ has been pre-processed as explained in Sect. 2 and has been segmented in frames with length of N samples. Each speech segment obtained to this end, represented by sðnÞ, n ¼ 0; 1; :::; N À 1, which was pre-emphasized and weighted by the Hamming window, is subject to the DFT,  Àj2pnk SðkÞ ¼ sðnÞ Á exp ; N n¼0 N À1 X  k ¼ 0; 1; :::; N À 1; where k is the index of the Fourier coefficients, SðkÞ.

The computation of the MFCC-FB40 can be summarized as follows. Let us denote with n the discrete-time index, and with xðnÞ a discrete-time speech signal that has been sampled with sampling frequency fs . Let us consider that the signal xðnÞ has been pre-processed as explained in Sect. 2 and has been segmented in frames with length of N samples. 28) 3 DFT-Based Speech Parameterization 35 Here, n is the index of the time-domain samples, and k is the index of the Fourier coefficients SðkÞ. 25) is employed in the computation of the log-energy output: Si ¼ log10 N À1 X !

2), and covering the frequency range [0, 5000] Hz. Thus, in the following, we refer to this implementation as to MFCC-FB20. The center frequencies of the first ten filters, residing in the frequency range [100, 1000] Hz, are linearly spaced, and the next ten have center frequencies logarithmically spaced between 1000 and 4000 Hz. 6) f ci ¼ 0:2ðiÀ10Þ fc10 Á 2 ; i ¼ 11; :::; 20 where the center frequency fci is assumed in Hz. 3 DFT-Based Speech Parameterization 25 Fig. 3 Mel-spaced filter-bank of equal-height filter according to Davis and Mermelstein (1980).

Download PDF sample

Rated 4.55 of 5 – based on 47 votes