Noisy speech recognition failure diagnosis using Minimum Message Length decision trees

Author

Cernak, Milos ; Darjaa, Sakhia

Author_Institution

Inst. of Inf., Slovak Acad. of Sci., Bratislava

fYear

2008

fDate

25-28 June 2008

Firstpage

409

Lastpage

412

Abstract

Current ASR technology lacks of effective failure diagnosis of ASR systems. Figures of merits such as WER are very useful, but donpsilat bring much insight into error patterns, error predictions or error analysis of ASR output. This paper explores an application of minimum message length (MML) style decision trees for such a diagnosis, focusing on theoretical background and the failure diagnosis of noisy speech recognition. In addition, the paper focuses on failure diagnosis of noisy speech, covering several kinds of intrinsic speech variabilities as well. Results on added speech-shaped noise at different SNR, ranging from 25 dB to -10 dB, are presented.

Keywords

decision trees; fault diagnosis; speech recognition; system recovery; error analysis; minimum message length; minimum message length decision trees; speech recognition failure diagnosis; Automatic speech recognition; Decision trees; Electronic mail; Encoding; Humans; Informatics; Loudspeakers; Machine learning; Speech analysis; Speech recognition; Automatic speech recognition; decision trees; failure diagnosis; machine learning;

fLanguage

English

Publisher

ieee

Conference_Titel

Systems, Signals and Image Processing, 2008. IWSSIP 2008. 15th International Conference on

Conference_Location

Bratislava

Print_ISBN

978-80-227-2856-0

Electronic_ISBN

978-80-227-2880-5

Type

conf

DOI

10.1109/IWSSIP.2008.4604453

Filename

4604453