Skip to main content

Table 5 Table 5

From: Towards an Intelligent Acoustic Front End for Automatic Speech Recognition: Built-in Speaker Normalization

1G:

Single-Gaussian

ACE:

Arcon communicability exercise

APT:

All-pass transform

ASR:

Automatic speech recognition

BISN:

Built-in speaker normalization

BTS:

Binary tree search

BLT:

Bilinear transform

CDHMM:

Continuous density hidden Markov model

CMN:

Cepstral mean normalization

DRT:

Diagnostic rhyme test

FFT:

Fast Fourier transform

GMM:

Gaussian mixture model

HMM:

Hidden Markov model

IFFT:

Inverse fast Fourier transform

LDA:

Linear discriminant analysis

LP:

Linear prediction

LPC:

Linear predictive coding

LPCCs:

Linear prediction-based cepstral coefficients

MFCCs:

Mel-frequency cepstral coefficients

MS-BTS:

Model space binary tree search

MVDR:

Minimum variance distortionless response

NFE:

Number of feature extraction

NLC:

Number of likelihood computation

NRPs:

Number of recognition passes

NWs:

Number of warps

PMVDR:

Perceptual MVDR cepstral coefficients

SNW:

Self-normalization warp

SPINEs:

Speech in noise evaluations

TM:

Trace measure

VAD:

Voice activity detector

VTLN:

Vocal-tract length normalization

VTTF:

Vocal-tract transfer function

WER:

Word error rate.