Abstract: A euclidean distance between an input pattern and a reference pattern is elementwise determined, to sequentially calculate a set of cumulative values thereof, which is stored in a work memory for use in a current recurrence calculation and in a cumulative data memory for use in a decision for a speech recognition. The recurrence calculation is executed on one of division blocks obtained by collecting input frames and reference frames. Stored data in the work memory are shifted every block for a subsequent recurrence calculation.
Abstract: A method for segmenting audio data, comprising speech from a plurality of individual speakers, according to speaker is provided. The method comprises providing individual HMMs for each individual speaker, each individual HMM including at least one state, and constructing a speaker network HMM by connecting the individual HMMs in parallel. The audio data is then divided into segments by determining a most likely sequence of states through the speaker network HMM, each of the segments being associated with one of the individual HMMs. Afterward, the speaker of each of the segments is identified. The segmented data may be used to form an index into the audio data according to speaker.
Type:
Grant
Filed:
April 12, 1994
Date of Patent:
August 5, 1997
Assignee:
Xerox Corporation
Inventors:
Vijay Balasubramanian, Francine R. Chen, Philip A. Chou, Donald G. Kimber, Alex D. Poon, Karon A. Weber, Lynn D. Wilcox
Abstract: A system for reconstructing a signal waveform from a correlogram is based upon the recognition that the information in each channel of the correlogram is equivalent to the magnitude of the Fourier transform of a signal. By estimating a signal on the basis of its Short-Time Fourier Transform Magnitude, each channel of information from a cochlear model can be reconstructed. Once this information is retrieved, a signal waveform can be resynthesized through inversion of the cochlear model. The process for reconstructing the cochlear model data can be optimized with the use of techniques for improving the initial estimate of the signal from the magnitude of its Fourier Transform, and by employing information that is known apriori about the signal during the estimation process, such as the characteristics of sound signals.
Type:
Grant
Filed:
February 22, 1993
Date of Patent:
December 5, 1995
Assignee:
Apple Computer, Inc.
Inventors:
Malcolm Slaney, Richard F. Lyon, Daniel Naar