Abstract :
Speech recognition for smartphones and other applications could be improved thanks to a new background noise estimation technique from researchers in Korea. Gil-Jin Jang from Ulsan National Institute of Science and Technology (UNIST) and Hong-Young Cho from the Electronics and Telecommunications Research Institute (ETRI) have created a technique to improve the quality of speech recognition. By tracking peak frequencies using line spectral pairs (LSPs) their method gives a more reliable and predictable noise classification performance, even in severe noise conditions.
