|
Abstract. In future command and control (C2) centres, staff should be
able to move about freely, unconstrained by microphone headsets, while
their conversations are transcribed to text using speaker-independent speech-recognition devices.
The output of the transcriber would be automatically labelled with
the staff’s identities. Automatic Speaker Identification (ASI) is a candidate
to perform that back-end function. However, ASI accuracy remains lower
than that of human speaker recognition, despite 40 years of
R&D, because the technology does not cope well with a
minority of speakers. In addition, the acoustical environment of C2
centres is very complex. It is strongly affected by reverberation
and the cocktail-party and Lombard effects. These and the presence
of low-energy speech degrade both speech and speaker recognition. Nevertheless,
that degradation is expected to vary significantly from centre to
centre. It is possible that the use of microphone arrays
can render ASI operational, in at least some of those
centres, especially if the number of speakers considered is small.
Related topics:
speaker identification, command systems
View first page of "Chatelain: Automatic Speaker Identification in C2 Centres: Challenges and Pitfalls"
Papers by Chatelain
Register for the free
to receive a list of papers for each issue as it is released.
|
|
