Data processing: speech signal processing – linguistics – language – Audio signal time compression or expansion
Reexamination Certificate
1998-03-10
2001-06-19
Knepper, David D. (Department: 2748)
Data processing: speech signal processing, linguistics, language
Audio signal time compression or expansion
C704S211000, C708S290000
Reexamination Certificate
active
06249766
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to processing digital data and more particularly to real time format conversion of digital audio waveform data.
2. Description of the Prior Art
As computers have become increasingly integrated into our culture, they have become intertwined with several existing technologies dealing with audio media. Computers are already prominent, or are becoming prominent, in telephony systems, radio systems, and speech interfaces to many types of devices. As a result, digital audio data has become much more common, and processing it efficiently has become an important issue.
An important problem that faces digital audio applications is that many of the subsystems from which such applications are constructed operate on different audio data formats. Although audio format conversion is a well-understood area, most conversions are accomplished off-line, with an emphasis on highly accurate conversion rather than on conversion speed. In modern digital audio systems, where many audio sources are real-time and produce transient data, off-line format conversion is not always acceptable. Some systems require “on the fly” format conversion, with the process completing within real-time constraints.
The traditional technique for down-sampling digital waveform data is described in various well-known sources, such as Oppenheim, A. and Schafer, R.,
Discrete-Time Signal Processing,
Prentice-Hall, 1989, p.101-112. This technique involves creating a discrete-time Fourier transform model of the audio signal and operating on it. Such a mechanism is favorable when a highly faithful down-sampling is required, but can be quite slow. In order to speed the process up to real-time speeds, a Fourier model with very few terms must be used. Although this may be acceptable for certain highly tonal (or cyclical) data sets, Fourier models with few terms are inaccurate models of speech and other complex waveforms. Thus, in the traditional system, the number of terms in the model provides a trade-off between speed and accuracy, and at the speeds required for real-time conversion, the accuracy becomes unacceptable for many types of data.
SUMMARY OF THE INVENTION
The present invention is a new down-sampling system for digital waveforms. The system is fast enough to use in real-time, “on the fly” conversions and results in data of acceptable quality for many applications, including applications dealing primarily with speech data.
Typically, the down-sampler of the present invention is located between an digital waveform producer and a digital waveform consumer. The down-sampler receives an input digital audio stream from the audio data producer and down-samples the data as it arrives. The output of the down-sampler is a down-sampled digital audio stream.
The down-sampler comprises a weight matrix calculator where a weights matrix needed for the down-sampling is calculated. Next a loop begins in which the system takes the input data from the producer's data stream, and at one chunk at a time, the system generates the output data. The loop comprises an input receiver, a chunk receiver, an output chunk generator, a chunk decider for deciding whether there is another chunk in the input, and an input decider for deciding whether there is more input. If there is not more input, the conversion is completed and the down-sampler of the present invention terminates. The generation of the weights matrix and the generation of the output data are critical parts of the invention.
REFERENCES:
patent: 5111505 (1992-05-01), Kitoh et al.
patent: 5341432 (1994-08-01), Suzuki et al.
patent: 5398029 (1995-03-01), Toyama et al.
patent: 5453741 (1995-09-01), Iwata
patent: 5621404 (1997-04-01), Heiss et al.
Goose Stuart
Wynblatt Michael J.
Knepper David D.
Paschburg Donald B.
Siemens Corporate Research Inc.
LandOfFree
Real-time down-sampling system for digital audio waveform data does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Real-time down-sampling system for digital audio waveform data, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Real-time down-sampling system for digital audio waveform data will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2482370