Filtering and Compression of Mel Sub-Band Energies for Speech Recognition in Noise
Publish place: 15th Iranian Conference on Electric Engineering
Publish Year: 1386
Type: Conference paper
Language: English
View: 2,258
This Paper With 6 Page And PDF Format Ready To Download
- Certificate
- I'm the author of the paper
Export:
Document National Code:
ICEE15_132
Index date: 6 February 2007
Filtering and Compression of Mel Sub-Band Energies for Speech Recognition in Noise abstract
The Mel-frequency cepstral cofficients (MFCC) are commonly used in speech recognition systems. Bu they are high sensitive to presence of
external noise. In this paper, we propose a noise compensation method for Mel filter bank energies and so MFCC features. This compensation method is performed in two stages: Mel sub-band filtering and then compression of Mel-sub-band energies. In the compression step, we propose a sub-bqnd SNRdependent compression function. We use this function in place of logarithm function in conventional MFCC feature extraction in presence of additive noise. Results show that the proposed nethod significantly improves MFCC features performance in noisy conditions where it decreases average word error rate up to 3094 for isolated word recognition on three test sets of Aurora 2 database.
Filtering and Compression of Mel Sub-Band Energies for Speech Recognition in Noise Keywords:
Filtering and Compression of Mel Sub-Band Energies for Speech Recognition in Noise authors
Babak Nasersharif
Computer Engineering Department, Iran University of Science and Technology
Ahmad Akbari
Computer Engineering Department, Iran University of Science and Technology
Mohantnmd Mehdi Honzayouttpour
Computer Engineering and IT department, Amirkabir University of Technology
مراجع و منابع این Paper:
لیست زیر مراجع و منابع استفاده شده در این Paper را نمایش می دهد. این مراجع به صورت کاملا ماشینی و بر اساس هوش مصنوعی استخراج شده اند و لذا ممکن است دارای اشکالاتی باشند که به مرور زمان دقت استخراج این محتوا افزایش می یابد. مراجعی که مقالات مربوط به آنها در سیویلیکا نمایه شده و پیدا شده اند، به خود Paper لینک شده اند :