Method and device for robust feature extraction of speech recognition
A speech recognition and feature extraction technology, applied in speech recognition, speech analysis, instruments, etc., can solve problems such as the inability to determine the short-term spectrum, achieve good speech recognition, and improve the effect of speech recognition
Inactive Publication Date: 2005-01-19
TELEFON AB LM ERICSSON (PUBL)
View PDF0 Cites 0 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Therefore, in this case, in those spectral regions where the spectral magnitude of the speech is equal to or less than the noise value, it cannot be determined whether there is a short-term spectrum containing speech
Method used
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View moreImage
Smart Image Click on the blue labels to locate them in the text.
Smart ImageViewing Examples
Examples
Experimental program
Comparison scheme
Effect test
Embodiment Construction
[0022] To make speech recognition more robust against noise, a robust feature extraction scheme can be employed. This scheme attempts to extract those features from the speech signal that are not sensitive to noise or are not affected by noise. Furthermore, this feature extraction scheme is mainly based on short-term spectrum analysis. Additionally, most speech recognition systems are based on short-term analysis in the MEL frequency range. The MEL frequency range is based on the human hearing range and is well known in the art, so it need not be described in depth here.
[0023] The term robustness shall include robustness to both stable and unstable background noise in the prior art mentioned above. In this application, in addition to the robustness mentioned above, the robustness to unknown frequency characteristics produced by any type of electronic equipment, such as the microphone and / or digital or is the frequency characteristic of the analog filter.
[0024] The fo...
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More PUM
Login to View More Abstract
The present invention relates to a method and an apparatus for a robust feature extraction for speech recognition in a noisy environment, wherein the speech signal is segmented and is characterized by spectral components. The speech signal is splitted into a number of short term spectral components in L subbands, with L = 1,2,...and a noise spectrum from segments that only contain noise is estimated. Then a spectral subtraction of the estimated noise spectrum from the corresponding short term spectrum is performed and a probability for each short term spectrum component to contain noise is calculated. Finally these spectral component of each short-term spectrum, having a low probability to contain speech are interpolated in order to smooth those short-term spectra that only contain noise. With the interpolation the spectral components containing noise are interpolated by reliable spectral speech components that could be found in the neighborhood.
Description
technical field [0001] The present invention relates to methods and apparatus for performing robust feature extraction for speech recognition in noisy environments. Background technique [0002] A major problem in the field of speech recognition is how to accurately recognize speech in noisy environments. All possible noises of different types can affect speech recognition and can cause a drastic deterioration in recognition accuracy. [0003] Especially in the field of mobile phones or access systems which allow access after recognition of a voice password, speech recognition becomes more important. Especially in these fields mentioned above, among the possible different types of noise, the most problematic is the additional stable or unstable background noise. Another type of noise that deteriorates the recognition accuracy is that when the voice to be recognized is sent through the transmission channel, it will be affected by the frequency characteristics of the transmi...
Claims
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More Application Information
Patent Timeline
Login to View More Patent Type & Authority Patents(China)
IPC IPC(8): G10L15/02G10L15/20G10L21/0208
CPCG10L21/0208G10L15/02G10L15/20
Inventor R·布吕克纳H·-G·希尔施R·克利施V·斯普林格
Owner TELEFON AB LM ERICSSON (PUBL)