Name: MELANIE

Text: MELANIE
Speaker Identification
GMM

CONFIDENTIAL
– Subject to–change
prior
notice!prior notice!
CONFIDENTIAL
Subjectwithout
to change
without

1

MELANIE – Speaker Identification, GMM

Functionality
Speaker identification determines the
probability of a set of speakers in speech
signals.
The algorithm is especially useful when
searching for a certain speaker in mass
audio data.

Training of Models
Goal
Training Data

Training is performed in order to optimise the models on the
respective speaker.
The training data must be labeled according to the prevailing
speaker. If possible, the speaker data should be obtained
from different channels and languages. A minimum of two
minutes speech is required for each speaker.

Technical Data
OperatingSystem Available on Windows XP and Linux
Processing Time

• On a P4; 3GHz, 1GB RAM computer
• 1/8 of real time; i.e. a signal of 8 minutes length is
processed in one minute

Recognition Rate The recognition rate depends on the number of speakers, on
the similarities of the speakers among each other and on the
speech quality.

Products
Mel-Pr-GMM-site
Mel-Pr-GMM-PC
Mel-Tr-GMM

Production environment for MELANIE GMM (site license)
Production environment for MELANIE GMM (PC license)
Training environment for MELANIE GMM

CONFIDENTIAL – Subject to change without prior notice!

2

Consequently, this means that a recording with a signal length of 20
minutes can be fully classified in one minute.

CONFIDENTIAL – Subject to change without prior notice!

3

Document Path: ["melanie-speaker-identification-gmm.pdf"]

e-Highlighter

Click to send permalink to address bar, or right-click to copy permalink.

Un-highlight all Un-highlight selectionu Highlight selectionh