Spectral Subtraction Using Model-based Spectra
Blekinge Institute of Technology, Department of Telecommunications and Signal Processing1999 (English)Conference paper (Refereed) Published
When using a mobile phone with a handsfree equipment in a vehicle it is difficult to have a relaxed conversation due to the high noise level. The noisy environment gives raise to a poor signal to noise ratio since the handsfree microphone picks up the background noise as well as the speech. By using speech enhancement processing, the quality of the conversation will be enhanced. This paper introduces a speech enhancement algorithm based on the spectral subtraction method. The method uses estimates of the noise spectrum and the noisy speech spectrum to form an SNR-based gain function, which is used for filtering the microphone signal. There are two different classes of spectrum estimation techniques non-parametric and parametric. Parametric estimation methods are based on a model of the underlying signal in order to estimate the spectrum.Furthermore a tool for assessing the accuracy of the estimations in situations when few data is available is applied, known as the bootstrapmethod. The spectral subtraction method, combined with different spectrum estimation techniques, are evaluated using real-world signals recorded in a car.
Place, publisher, year, edition, pages
Lyngby, Denmark, 1999.
IdentifiersURN: urn:nbn:se:bth-9703Local ID: oai:bth.se:forskinfoA446A4572FC2ADF5C1256C6A004536FFOAI: oai:DiVA.org:bth-9703DiVA: diva2:837623
the Sixth International Congress on Sound and Vibration