Loading...
Please wait, while we are loading the content...
Similar Documents
Robust speech end-point detection and f0 extraction by applying image processing techniques 1.
| Content Provider | CiteSeerX |
|---|---|
| Author | Bo, Zhang Gang, Peng |
| Abstract | This paper addresses two problems of robust speech recognition in noisy environment, i.e., robust speech endpoint detection and F0 extraction. Conventional speech end-point detectors base on the features such as signal energy, zero crossing rate, energy of linear prediction error, etc. By observing more than 130 kinds of realworld environmental noises, we may believe that the essential features of speech signal may be its harmonic structure and its formant structure. An end-point detector based on detecting valid speech harmonic structure was designed and studied in this research. It operates on speech spectrogram by making use of some image processing techniques. The detector first extracts speech harmonic structure from the noisy speech signal. Then the F0 contour is calculated from the harmonic structure. Finally, the speech end-point can be deduced from the boundary of the F0 contour. The algorithm shows good robustness when tested on various noisy speech signals. 1. |
| File Format | |
| Access Restriction | Open |
| Subject Keyword | F0 Contour Harmonic Structure Good Robustness Speech Harmonic Structure Robust Speech Recognition Noisy Environment Noisy Speech Signal End-point Detector Essential Feature Signal Energy Robust Speech Endpoint Detection Valid Speech Harmonic Structure Speech Signal Conventional Speech End-point Detector Base Formant Structure Various Noisy Speech Signal Realworld Environmental Noise Linear Prediction Error F0 Extraction Speech End-point Image Processing Technique Speech Spectrogram |
| Content Type | Text |
| Resource Type | Article |
