Liquid Audio, Inc.

Microsoft Technology Licensing, LLC

Neurotone, Inc.

Earl I Levine

Jason S Brownell

Steven T Ansell

Perry T Devine

Philip R Wiser

A Robert Modeste

Phil Chou

Gerald W Kearby

A hearing application suite includes enhancement and training for listening and hearing of prerecorded speech, extemporaneous voice communication, and non-speech sound. Enhancement includes modification of audio according to audiometric data representing subjective hearing abilities of the user, display of textual captions contemporaneously with the display of the audiovisual content, user-initiated repeating of a most recently played portion of the audiovisual content, user-controlled adjustment of the rate of playback of the audiovisual content, user-controlled dynamic range compression/expansion, and user controlled noise reduction. Training includes testing the user's ability to discern speech and/or various other qualities of audio with varying degrees of quality.

Personal hearing suite

To provide improved security in adjunct program modules such as plug-ins and dynamic link libraries, a requesting module provides an authorization interface to the invoked module such that the invoked module can require a certificate of the requesting module and can also challenge the authority of the requesting module. The certificate can include one or more permissions which are prerequisites for processing by the invoked module. The invoked module can challenge the authority of the requesting module by sending random test data to the requesting module and receiving in response a cryptographic signature of the test data. By verifying the signature of the requesting module using the received certificate, the invoked module confirms that the requesting module is, in fact, the owner of the receive certificate.

Security mechanism for computer processing

Security mechanism for computer processing modules

A digitized signal is published with a blank watermark, i.e., a watermark which contains no specific watermark data, and a server computer system encodes specific watermark data into the watermark signal for each delivery of the digitized signal. A basis signal which is used to watermark a digitized signal is predetermined to enable embedding of transaction-specific watermark data to be embedded in the subject signal with minimal processing resources. Since the basis signal generally accounts for the majority of the processing resources required to watermark the subject signal, only a relatively small portion of the requisite processing resources are used to embed transaction-specific watermark into the subject signal.

Server-side watermark data writing method and apparatus for digital signals

Watermark data is encoded in a digitized signal by forming a noise threshold spectrum which represents a maximum amount of imperceptible noise, spread-spectrum chipping the noise threshold spectrum with a relatively endless stream of pseudo-random bits to form a basis signal, dividing the basis signal into segments, and filtering the segments to smooth segment boundaries. The data encoded in the watermark signal is precoded to make the watermark data inversion robust and is convolutional encoded to further increase the likelihood that the watermark data will subsequently be retrievable notwithstanding lossy processing of the watermarked signal.

Robust watermark method and apparatus for digital signals

Watermark data is encoded in a digitized signal by forming a noise threshold spectrum which represents a maximum amount of imperceptible noise, spread-spectrum chipping the noise threshold spectrum with a relatively endless stream of pseudo-random bits to form a basis signal, dividing the basis signal into segments, and filtering the segments to smooth segment boundaries. The data encoded in the watermark signal is precoded to make the watermark data inversion robust and is convolutional encoded to further increase the likelihood that the watermark data will subsequently be retrievable notwithstanding lossy processing of the watermarked signal. A watermark alignment module determines which of a large number of offsets of the watermarked data is most likely to correspond to a recognizable watermark. The watermark alignment module uses a single basis signal to evaluate a number of offsets over a relatively narrow range of offsets. In addition, offsets which differ by an integer multiple of a spatial/temporal granularity of respective noise threshold spectra are recognized as corresponding to equivalent noise threshold spectra. Accordingly, a previously generated noise threshold spectrum for one offset is reused for a second offset which differs by an integer multiple of the spatial/temporal granularity.

Determination of a best offset to detect an embedded pattern

Watermark data is encoded in a digitized signal by forming a noise threshold spectrum which represents a maximum amount of imperceptible noise, spread-spectrum chipping the noise threshold spectrum with a relatively endless stream of pseudo-random bits to form a basis signal, dividing the basis signal into segments, and filtering the segments to smooth segment boundaries. The data encoded in the watermark signal is precoded to make the watermark data inversion robust and is convolutional encoded to further increase the likelihood that the watermark data will subsequently be retrievable notwithstanding lossy processing of the watermarked signal. To produce the endless pseudo-random bit stream, subsequent bits of the sequence are generated in a pseudo-random manner from previous bits of the sequence. The pseudo-random bits are appended to the stream of pseudo-random bits and, additionally, replace a number of bits of the state. The remaining bits are shifted such that, in effect, the least recently generated bits of the state are discarded. Thus, after production of a number of pseudo-random bits of the stream, the state changes significantly. Accordingly, repeating the above process generates additional pseudo-random bits which are very likely to differ significantly from the previously generated pseudo-random bits. A basis signal is generated by spread-spectrum chipping in accordance with the stream of pseudo-random bits and watermark data is encoded in the basis signal to form a watermark signal which is added to a digitized signal.

Secure watermark method and apparatus for digital signals

Watermark data is encoded in a digitized signal by forming a noise threshold spectrum which represents a maximum amount of imperceptible noise, spread-spectrum chipping the noise threshold spectrum with a relatively endless stream of pseudo-random bits to form a basis signal, dividing the basis signal into segments, and filtering the segments to smooth segment boundaries. The data encoded in the watermark signal is precoded to make the watermark data inversion robust and is convolutional encoded to further increase the likelihood that the watermark data will subsequently be retrievable notwithstanding lossy processing of the watermarked signal. The basis signal fits noise thresholds determined by constant-quality quantization approximation. Noise introduced by quantization is estimated by determining a continuously differentiable function which approximates noise introduced by such quantization and using the function to solve for a relatively optimal gain to be applied during such quantization. The continuously differentiable function includes a local quantization stepsize. A local quantization stepsize is determined by first determining widths of quantization steps at respective particular amplitudes and interpolating stepssizes for amplitudes between the particular amplitudes. The interpolation of stepsizes provides a smooth function. The continuously differentiable function based upon a local, interpolated quantization stepsize provides an estimation of quantization error which lends itself to efficient and convenient mathematical manipulation.

Efficient watermark method and apparatus for digital signals

Watermark data is encoded in a digitized signal by forming a noise threshold spectrum which represents a maximum amount of imperceptible noise, spread-spectrum chipping the noise threshold spectrum with a relatively endless stream of pseudo-random bits to form a basis signal, dividing the basis signal into segments, and filtering the segments to smooth segment boundaries. The data encoded in the watermark signal is precoded to make the watermark data inversion robust and is convolutional encoded to further increase the likelihood that the watermark data will subsequently be retrievable notwithstanding lossy processing of the watermarked signal. Watermark data is encoded in a basis signal by division of the basis signal into segments and inverting the basis signal in segments corresponding to watermark data bits with a first logical value and not inverting the basis signal in segment corresponding to watermark data bits with a different logical value. The basis signal is smoothed at segment boundaries to eliminate any such discontinuities. Good results are achieved when scaling the basis signal by a cube-root of the positive half of a sine function which is aligned with segment boundaries such that the cube-root sine function tapers to zero at segment boundaries.

An adaptive linear predictor is used to predict samples, and residuals from such predictions are encoded using Golomb-Rice encoding. Linear prediction of samples of a signal which represents digitized sound tends to produce relatively low residuals and those residuals tend to be distributed exponentially. Accordingly, linear prediction combined with Golomb-Rice encoding produces particularly good compression rates with very efficient and simple implementation. The accuracy of the linear predictor is improved by including, in the prediction of a current sample of a first channel of the digitized signal, look-ahead sample data from a corresponding second channel of the digitized signal. For example, prediction of a right channel sample of a digitized, stereo, audio signal is improved by inclusion of look-ahead left channel sample data in the right channel sample predictor. Since correlated channels of a digitized signal have a relatively high degree of correlation, accuracy of sample prediction is improved and residuals between the predicted samples and actual samples are reduced, thereby reducing the amount of data required to represent the residuals. In reconstructing the encoded samples, a decoder uses a current reconstructed sample of the first channel of a digitized signal to predict a current sample of the second corresponding channel.

Lossless data compression with low complexity

An adaptive linear predictor is used to predict samples, and residuals from such predictions are encoded using Golomb-Rice encoding. Linear prediction of samples of a signal which represents digitized sound tends to produce relatively low residuals and those residuals tend to be distributed exponentially. Accordingly, linear prediction combined with Golomb-Rice encoding produces particularly good compression rates with very efficient and simple implementation. A code length used in Golomb-Rice, which is typically referred to as the parameter k, is adapted for each sample in a predictable and repeatable manner to further reduce the size of a Golomb-Rice encoding for each sample. An infinite incident response filter of processed residuals automatically reduces influences of previously processed residuals upon such adaptation as additional samples are processed. In addition, the influence of each residual processed in the adaptation of the code length is directly related to the recency of the processing of the sample to which the residual corresponds. Furthermore, no resetting of the code length adaptation mechanism is required since influence of particularly distant residuals upon the adaptation of the code length diminishes to negligible amounts over time. The efficiency of Golomb-Rice encoding is improved by limiting the predicted samples to an efficient range. The maximum of the efficient range is the maximum valid value of a sample less the maximum positive value of the fixed-length, binary portion of an encoded residual. The adaptive predictor adapts to residuals between actual and predicted samples at a particular rate.

An adaptive linear predictor is used to predict samples, and residuals from such predictions are encoded using Golomb-Rice encoding. Linear prediction of samples of a signal which represents digitized sound tends to produce relatively low residuals and those residuals tend to be distributed exponentially. Accordingly, linear prediction combined with Golomb-Rice encoding produces particularly good compression rates with very efficient and simple implementation. A code length used in Golomb-Rice, which is typically referred to as the parameter k, is adapted for each sample in a predictable and repeatable manner to further reduce the size of a Golomb-Rice encoding for each sample. An infinite incident response filter of processed residuals automatically reduces influences of previously processed residuals upon such adaptation as additional samples are processed. The efficiency of Golomb-Rice encoding is improved by limiting the predicted samples to an efficient range. The maximum of the efficient range is the maximum valid value of a sample less the maximum positive value of the fixed-length, binary portion of an encoded residual. The minimum of the efficient range is the minimum valid value of a sample plus the minimum negative value of the fixed-length, binary portion of an encoded residual. Such reduces the number of bits required to represent a variable-length, unary portion of the encoded residual by improving the efficiency with which the fixed-length, binary portion can represent a particular residual.

An adaptive predictor is used to predict samples, and residuals from such predictions are encoded using Golomb-Rice encoding to thereby compress a digital signal which includes the samples. The adaptive predictor adapts to residuals between actual and predicted samples at a particular rate. The rate of adaptation is itself adapted periodically to ensure optimum performance of the compression. A number of samples are repeatedly compressed using different adaptation rates, and the adaptation rate which produces the best compression results is used. The adaptation rate can be an exponential power of 2 such that incrementing the adaptation rate effectively doubles the rate at which the predictor is adapted and decrementing the adaptation rate effectively halves the rate at which the predictor is adapted. Accordingly, fewer trials are needed to find a relatively optimal adaptation rate for the predictor. A code length used in Golomb-Rice, which is typically referred to as the parameter k, is adapted for each sample in a predictable and repeatable manner to further reduce the size of a Golomb-Rice encoding for each sample. An infinite incident response filter of processed residuals automatically reduces influences of previously processed residuals upon such adaptation as additional samples are processed. The efficiency of Golomb-Rice encoding is improved by limiting the predicted samples to an efficient range.

Growing community of inventors

Palo Alto, CA, United States of America

Earl I Levine