2006 IEEE International Conference on Acoustics Speed and Signal Processing Proceedings
DOI: 10.1109/icassp.2006.1660588
|View full text |Cite
|
Sign up to set email alerts
|

The Effect of Memory Inclusion on Mutual Information Between Speech Frequency Bands

Abstract: In this paper, we investigate the effect of temporal correlation on the dependence between the speech narrow and high frequency bands covering the 0.3-3.4 kHz and 3.7-8 kHz ranges, respectively. We follow the technique of using Gaussian mixture modelling of spectral envelopes represented by Mel-frequency cepstral coefficients. The correlation between the disjoint speech frequency bands is quantified through mutual information (MI) and its ratio to highband entropy. Speech exhibits considerable temporal correla… Show more

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
4
1

Citation Types

2
15
0

Publication Types

Select...
7
1

Relationship

0
8

Authors

Journals

citations
Cited by 12 publications
(17 citation statements)
references
References 10 publications
2
15
0
Order By: Relevance
“…[4,12,13], capture memory in the front-end instead, e.g., via delta features or static features from neighbouring frames. Following an investigation of front-end feature extraction for ABE [14], the work in [15][16][17] investigates the merit of memory inclusion through information theoretic analysis. This body of work demonstrates the benefit of memory inclusion through delta features under the constraint of fixed dimensionality.…”
Section: Introductionmentioning
confidence: 99%
“…[4,12,13], capture memory in the front-end instead, e.g., via delta features or static features from neighbouring frames. Following an investigation of front-end feature extraction for ABE [14], the work in [15][16][17] investigates the merit of memory inclusion through information theoretic analysis. This body of work demonstrates the benefit of memory inclusion through delta features under the constraint of fixed dimensionality.…”
Section: Introductionmentioning
confidence: 99%
“…In our recent work [3], we exploited the considerable temporal correlation properties of speech by including memory in Mel Frequency Cepstral Coefficient (MFCC) speech parametrization (through delta features). These features are obtained through linearly weighted differences between neighbouring conventional static feature vectors.…”
Section: Introductionmentioning
confidence: 99%
“…As described in [1], we include memory directly in the representation of spectral envelopes by means of delta coefficients, which are appended to (or replace part of) the MFCC/LSF static features 4 . Delta coefficients are obtained from static vectors by a first-order regression (time-derivative) implemented by calculating linearly weighted differences between neighbouring static vectors per…”
Section: Delta Featuresmentioning
confidence: 99%
“…In our previous work; first introduced in [1] and later extended in [2], we made use of the concept of highband certainty (certainty about the high band given the narrow band); defined in [3] as the ratio of Mutual Information (MI) between the two bands to the discrete entropy of the high band, in order to quantify the correlation between speech frequency bands. Through highband certainty, we investigated the effect of including memory into the frontend on the resulting correlation (by using delta features in addition to the conventional static features which make no use of the considerable temporal correlation properties of speech), as well as the effect of the type of parametrization.…”
Section: Introductionmentioning
confidence: 99%