Abstract: Disclosed herein are a stereo extension apparatus and method. The apparatus includes a database that stores predetermined information as a result of Gaussian mixture model (GMM) training or hidden Markov model (HMM) training; a modified discrete cosine transform (MDCT) transformer that transforms a mono signal through MDCT, a feature parameter extractor that extracts a feature parameter of the mono signal from an MDCT coefficient output from the MDCT transformer, a side signal energy estimator that estimates subband energy of a side signal with reference to information stored in the database based on the feature parameter; an energy controller that obtains the MDCT coefficient of a side signal estimated from the subband energy of the estimated side signal, an inverse MDCT transformer that obtains an estimated side signal by transforming the MDCT coefficient of the estimated side signal through inverse MDCT.

Abstract: Disclosed is a speech receiving apparatus. A low-band PLC module and a synthesis filter reconstructs a low-band speech signal of a lost frame from a previous good frame. A high-band PLC module reconstructs a high-band speech signal of the lost frame from the previous good frame. A transforming part transforms the low-band speech signal into a frequency range. A bandwidth extending part generates at least an extended MDCT coefficient as information for the high-band speech signal from the low-band speech signal transformed by the transforming part. A smoothing part smoothes the extended MDCT coefficient. An inverse transforming part inversely transforms the extended MDCT coefficient smoothed by the smoothing part to a time domain. A synthesizing part synthesizes the low-band speech signal, and the high-band speech signal which is inverse-transformed by the inverse transforming part and reconstructed, to output a wideband speech signal.

Abstract: A method and apparatus for audio signal encoding for noise reduction are provided. The method includes: receiving an audio signal and performing modified discrete cosine transformation (MDCT) on the audio signal to convert the audio signal into a long block or a short block; reducing noise included in the audio signal in accordance with the long block or the short block; and performing advanced audio coding (AAC) on the long block or the short block in which noise is reduced.

Abstract: The present invention extracts azimuth information on a sound source, read the touch state of a touch screen on which an image is displayed, and enables a sound source having azimuth information corresponding to a place touched on the image to be synthesized so as to be distinguished from other sound sources. According to the present invention, since it is possible to listen to the distinguished sound of a desired location on an image, a user may be provided with more satisfaction.

Abstract: Provided are an apparatus and method for eliminating noise. The method includes: detecting a speech section from a noise speech signal including a noise signal; separating the speech section into a consonant section and a vowel section on the basis of a VOP at the speech section; calculating a transfer function of a filter for eliminating the noise signal to allow the degree of noise elimination to be different in the consonant section and the vowel section; and eliminating the noise signal from the noise speech signal on the basis of the transfer function.

Abstract: An apparatus for correcting errors in speech recognition is provided. The apparatus includes a feature vector extracting unit extracting feature vectors from a received speech. A speech recognizing unit recognizes the received speech as a word sequence on the basis of the extracted feature vectors. A phoneme weighted finite state transducer (WFST)-based converting unit converts the recognized word sequence recognized by the speech recognizing unit into a phoneme WFST. A speech recognition error correcting unit corrects errors in the converted phoneme WFST. The speech recognition error correcting unit includes a WFST synthesizing unit modeling a phoneme WFST transferred from the phoneme WFST-based converting unit as pronunciation variation on the basis of a Kullback-Leibler (KL) distance matrix.

Abstract: Disclosed is an apparatus for extending a bandwidth of a sound signal. The apparatus includes a database that stores predetermined training information as a result of at least one of Gaussian mixture model (GMM) training and hidden Markov model (HMM) training; a modified discrete cosine transform (MDCT) transformer that transforms a first band signal through MDCT, a feature extractor that extracts a feature parameter of the first band signal from an MDCT coefficient output from the MDCT transformer; an extender that provides an extended MDCT coefficient for a second band signal based on the MDCT coefficient of the first band signal output from the MDCT transformer, a subband energy estimator that estimates subband energy of the second band signal with reference to information stored in the database based on the feature parameter.

Abstract: Disclosed herein are a stereo extension apparatus and method. The apparatus includes a database that stores predetermined information as a result of Gaussian mixture model (GMM) training or hidden Markov model (HMM) training; a modified discrete cosine transform (MDCT) transformer that transforms a mono signal through MDCT, a feature parameter extractor that extracts a feature parameter of the mono signal from an MDCT coefficient output from the MDCT transformer, a side signal energy estimator that estimates subband energy of a side signal with reference to information stored in the database based on the feature parameter; an energy controller that obtains the MDCT coefficient of a side signal estimated from the subband energy of the estimated side signal, an inverse MDCT transformer that obtains an estimated side signal by transforming the MDCT coefficient of the estimated side signal through inverse MDCT.

Abstract: A method for extending a bandwidth of a speech signal received, according to an embodiment of the present invention, includes: transforming the received speech signal into a frequency domain by decoding the received speech signal; normalizing the transformed speech signal; differentiating a voiced sound period or unvoiced sound period from the received speech signal; extracting, from the normalized speech signal, a first period including a harmonic component of the voiced sound period on the basis of the voiced sound period; extracting, from the normalized speech signal, a second period on the basis of correlation between the unvoiced sound period and the normalized speech signal; generating a high-band speech signal on the basis of the first period and the second period; and synthesizing the generated high-band speech signal and the transformed speech signal to output a wideband speech signal.

Abstract: A frame erasure concealment technique for a bitstream-based feature extractors in a speech recognition system particularly suited for use in a wireless communication system operates to “delete” each frame in which an erasure is declared. The deletions thus reduce the length of the observation sequence, but have been found to provide for sufficient speech recognition based on both single word and “string” tests of the deletion technique.

Abstract: Provided is an apparatus and method for encoding a voice and audio signal by expanding a modified discrete cosine transform (MDCT) based CODEC to a wideband and a super-wideband in a communication system. The apparatus for encoding a signal in a communication system, includes a converter configured to convert a time domain signal corresponding to a service to be provided to users to a frequency domain signal, a quantization and normalization unit configured to calculate and quantize gain of each subband in the converted frequency domain signal and normalize a frequency coefficient of the each subband, a search unit configured to search patch information of each subband in the converted frequency domain signal using the normalized frequency coefficient, and a packetizer configured to packetize the quantized gain and the searched patch information and encode gain information of each subband in the frequency domain signal.

Abstract: A frame erasure concealment technique for a bitstream-based feature extractor in a speech recognition system particularly suited for use in a wireless communication system operates to “delete” each frame in which an erasure is declared. The deletions thus reduce the length of the observation sequence, but have been found to provide for sufficient speech recognition based on both single word and “string” tests of the deletion technique.

Abstract: Method and device of extending a signal band of a voice or audio signal are provided. The bandwidth extension method includes the steps of: performing a modified discrete cosine transform (MDCT) process on an input signal to generate a first transform signal; generating a second transform signal and a third transform signal on the basis of the first transform signal; generating normalized components and energy components of the first transform signal, the second transform signal, and the third transform signal therefrom; generating an extended normalized component from the normalized components and generating an extended energy component from the energy components; generating an extended transform signal on the basis of the extended normalized component and the extended energy component; and performing an inverse MDCT (IMDCT) process on the extended transform signal.

Abstract: An apparatus for correcting errors in speech recognition is provided. The apparatus includes a feature vector extracting unit extracting feature vectors from a received speech. A speech recognizing unit recognizes the received speech as a word sequence on the basis of the extracted feature vectors. A phoneme weighted finite state transducer (WFST)-based converting unit converts the recognized word sequence recognized by the speech recognizing unit into a phoneme WFST. A speech recognition error correcting unit corrects errors in the converted phoneme WFST. The speech recognition error correcting unit includes a WFST synthesizing unit modeling a phoneme WFST transferred from the phoneme WFST-based converting unit as pronunciation variation on the basis of a Kullback-Leibler (KL) distance matrix.

Abstract: Disclosed is a speech receiving apparatus. A low-band PLC module and a synthesis filter reconstructs a low-band speech signal of a lost frame from a previous good frame. A high-band PLC module reconstructs a high-band speech signal of the lost frame from the previous good frame. A transforming part transforms the low-band speech signal into a frequency range. A bandwidth extending part generates at least an extended MDCT coefficient as information for the high-band speech signal from the low-band speech signal transformed by the transforming part. A smoothing part smoothes the extended MDCT coefficient. An inverse transforming part inversely transforms the extended MDCT coefficient smoothed by the smoothing part to a time domain. A synthesizing part synthesizes the low-band speech signal, and the high-band speech signal which is inverse-transformed by the inverse transforming part and reconstructed, to output a wideband speech signal.

Abstract: A speech analysis apparatus is provided. An F0 extraction part extracts a pitch value from speech information. A spectrum extraction part extracts spectrum information from the speech information. An MVF extraction part extract a maximum voiced frequency and allows boundary information for respectively filtering a harmonic component and a non-harmonic component to be obtained. According to the speech analysis apparatus, speech synthesis apparatus, and speech analysis synthesis system of the present invention, speech that is closer to the original voice and is more natural may be synthesized. Also, speech may be represented with less data capacity.

Abstract: A method and apparatus for audio signal encoding for noise reduction are provided. The method includes: receiving an audio signal and performing modified discrete cosine transformation (MDCT) on the audio signal to convert the audio signal into a long block or a short block; reducing noise included in the audio signal in accordance with the long block or the short block; and performing advanced audio coding (AAC) on the long block or the short block in which noise is reduced.

Abstract: The invention comprises a method and apparatus for predicting word accuracy. Specifically, the method comprises obtaining an utterance in speech data where the utterance comprises an actual word string, processing the utterance for generating an interpretation of the actual word string, processing the utterance to identify at least one utterance frame, and predicting a word accuracy associated with the interpretation according to at least one stationary signal-to-noise ratio and at least one non-stationary signal to noise ratio, wherein the at least one stationary signal-to-noise ratio and the at least one non-stationary signal to noise ratio are determined according to a frame energy associated with each of the at least one utterance frame.

Abstract: The example embodiment of the present invention provides an acoustic model adaptation method for enhancing recognition performance for a non-native speaker's speech. In order to adapt acoustic models, first, pronunciation variations are examined by analyzing a non-native speaker's speech. Thereafter, based on variation pronunciation of a non-native speaker's speech, acoustic models are adapted in a state-tying step during a training process of acoustic models. When the present invention for adapting acoustic models and a conventional acoustic model adaptation scheme are combined, more-enhanced recognition performance can be obtained. The example embodiment of the present invention enhances recognition performance for a non-native speaker's speech while reducing the degradation of recognition performance for a native speaker's speech.

Abstract: Provided are an apparatus and method for eliminating noise. The method includes: detecting a speech section from a noise speech signal including a noise signal; separating the speech section into a consonant section and a vowel section on the basis of a VOP at the speech section; calculating a transfer function of a filter for eliminating the noise signal to allow the degree of noise elimination to be different in the consonant section and the vowel section; and eliminating the noise signal from the noise speech signal on the basis of the transfer function.