Abstract: Acoustic echos are canceled by employing a first echo canceler having a comparatively short first impulse response synthesis capability which is connected between a transmit path and receive path for generating a first error signal and for canceling echo signals in the transmit path, and at least a second echo canceler having a comparatively long second impulse response synthesis capability connected in parallel with the first echo canceler between the transmit and the receive path. The second echo canceler is supplied with the first error signal from the first echo canceler and is adaptively operating simultaneously with but independent of the first echo canceler to further cancel echos in the transmit path. More specifically, the first echo canceler is intended to capture the direct path acoustic echo and any early arriving echos that are stable which are not time yawing. Since the direct path acoustic echo and the stable early arriving echos are not time varying, the adaptive adaptation rate, i.e.

Abstract: Failure to cancel echos at frequencies in which the frequency characteristics of sub-band filters in the X and Y analysis filter banks overlag in subband echo cancelers is overcome by employing an X-analysis filter bank in which the pass-band of the subband filters is wider than the pass-band of the corresponding subband filters in the Y-analysis filter bank. This insures that there is sufficient energy in the tap delay lines of the adaptive filters in the subband echo cancelers at all frequencies where echos are to be synthesized. More specifically, the keeping of sufficient energy in the tap delay lines of the adaptive filters in the subband echo cancelers avoids exciting eigenmodes associated with the small eigenvalues of the correlation matrix.

Abstract: More efficent encoding/decoding in decomposing and recomposing a high resolution image is obtained by employing a unique pixel cluster arrangement. The pixel clusters of an image are classified as one of a plurality of classifications. A unique encoding/decoding technique is assigned to each classification. In one embodiment, a cluster including at least one high resolution pixel to be recomposed from a corresponding low resolution pixel which is a so-called exception to general prediction rules is classified as a first classification. A cluster including no pixels which are exceptions is classified as a second classification. Supplemental information is encoded/decoded for all pixels in a cluster classified as the first classification. Pixels in the cluster classified as the second classification are encoded/decoded according to the general prediction rules.

Abstract: In entropy, e.g. arithmetic, encoding and decoding, probability estimates are needed of symbols to be encoded and subsequently decoded. More accurate probability estimates are obtained by controllably adjusting the adaptation rate of an adaptive probability estimator. The adaptation rate is optimized by matching it to the actual probability values being estimated. In particular, the adaptation rate is optimized to be proportional to the inverse of the smallest value probability being estimated. Consequently, if the probability values being estimated are not small a "fast" adaption rate is realized and if the probability values being estimated are small a necessarily slower adaptation rate is realized.

Abstract: In entropy, e.g., arithmetic or adaptive Huffman, encoding/decoding a context based on prior symbols is needed to provide accurate predictions of symbols to be encoded/decoded. Encoding/decoding efficiency is improved by employing an adaptive context extractor. The adaptive context extractor automatically adjusts the configuration of the lag intervals used to define the context. This is realized by adaptively incorporating into the context configuration at least one lag interval found to have a "good" predictive value relative to the particular symbol stream being encoded/decoded. The context configuration adjustment is such that the at least one found lag interval is exchanged with the lag interval currently in an at least one so-called floating predictor position.

Abstract: More efficient encoding/decoding in decomposing and recomposing a high resolution image is obtained by employing a unique prediction arrangement. The prediction arrangement determines whether high resolution pixels to be recomposed from low resolution pixels from a low resolution replica of the high resolution are so-called typically predictable or non-typically predictable by using general prediction rules. The general prediction rules are also employed to determine if any of the typically predictable high resolution pixels would be improperly recomposed. Such high resolution pixels which would be improperly recompased are identified as exceptions. It is noted that so-called supplemental information is required to properly recompose non-typically predictable pixels and typically predictable pixels which are identified as exceptions. To this end, exceptions accompany the low resolution pixel for which the corresponding high resolution pixels would otherwise be improperly recomposed.

Abstract: In order to employ an output register having a finite number of stages in an arithmetic encoder, it is necessary to provide carryover control, otherwise a register having an impractically large number of stages would be required, i.e., a so-called "infinite" register. The so-called "infinite" output register is emulated by employing a counter and a finite register. To this end, a count is accumulated of sets, i.e., bytes, of consecutive prescribed logical signals of a first kind, i.e., logical 1's, being generated by an arithmetic coding register and possibly modified by a carry indication. The accumulated count is then employed to supply as an output a like number of sets including logical signals of a second kind, i.e., logical 0's, or logical signals of the first kind, i.e., logical 1's, depending on whether or not a carry would propagate through the stages of the so-called "infinite" register being emulated.

Abstract: A high resolution image is decomposed to a basic lower resolution replica and a plurality of supplemental information fields for use in a progressive high resolution image facsimile transmission system. The basic lower resolution replica is generated by successive decompositions of the original image into a plurality of intermediate lower resolution replicas and corresponding supplemental information fields by employing a so-called edge decomposition technique. The supplemental information fields are required in a receiver to upgrade the basic lower resolution replica toward the original high resolution replica.

Abstract: Impulse response estimation is improved in an adaptive filter by compensating digital signals incoming to the filter to include a nonlinearity substantially identical to a nonlinearity included in a return signal path to the filter.

Abstract: Update gain normalization is employed in adaptive filters to control weighting of the filter impulse response updates in order to converge properly to a desired impulse response. Singing, i.e., oscillating, of the filter is overcome by adjusting the update gain when an incoming signal power estimate used to normalize the gain exceeds a prescribed threshold value. In one example, the normalized gain is adjusted to be a fixed value for power estimate values which exceed the threshold. In accordance with another aspect of the invention, a single normalized gain value is used to adjust the update gain in two adaptive filters employed as echo cancelers in a bidirectional voice frequency repeater.

Abstract: An estimate of delay (.DELTA.) in a communication circuit is obtained by subsampling (via 110-114) signals incoming (X(K)) to and outgoing (Y(K)) from the communication circuit, obtaining a cross correlation (via 120, 121) of the subsampled signals and processing (via 122) the cross correlation in accordance with a prescribed algorithm.In one application (FIG. 1) the delay estimate (.DELTA.) is used in conjunction with an adjustable delay (101) connected in series with a "short" echo canceler (102) across the communication circuit for cancelling echoes.

Abstract: An energy discriminator is employed in conjunction with an adaptive filter to control updating of the filter transfer function characteristic. Specifically, the discriminator is employed to distinguish whether any significant received far end energy is only partial band or whole band. If the received energy is partial band the adaptive filter is inhibited from updating the transfer function characteristic during intervals that such energy is being received. On the other hand, if the received energy is determined not to be partial band and, hence, is whole band, the filter is enabled to update the transfer function characteristic during intervals that such energy is being received.In a specific example, the discriminator is employed in an adaptive echo canceler to inhibit updating an echo path estimate being generated by an adaptive transversal filter when partial band energy is being received and enabling updating of the echo path estimate when whole band energy is being received.

Abstract: Loop gain normalization is employed in adaptive filters to control weighting of the filter characteristic updates in order to converge properly to a desired filter characteristic. Filter stability and rapid high quality convergence is realized for a variety of received or inputted signals by employing both long term and fast attack estimates of a prescribed input signal characteristic to normalize the update gain. In one embodiment, both long term and fast attack input signal power estimates are generated and one of the two estimate values is selected to normalize the update gain. Specifically, the fast attack estimate is modified by a predetermined value and, then, the larger of the long term estimate and modified fast attack estimate is selected to normalize the update gain.

Abstract: Adaptive filters are commonly used in echo cancelers and automatic equalizers. Usually adaptive filters include a tapped delay line and apparatus coupled to the delay line for producing a tap coefficient signal, whose sign and magnitude indicate the appropriate correction in adjusting the filter. However, in the presence of input signals having a partial frequency band spectrum, known filters tend to become unstable, e.g., tap coefficient signals blow up. The instant arrangement includes apparatus for weakly driving the tap coefficient signals to optimal values. As illustrated in a deceptively simple embodiment, a tap coefficient updating component is extended through a one's complement converter to a first input of a binary adder. A two's complement output of the adder is fed back to a second input of the adder. The sign of the adder output is also provided to a CARRY-IN input terminal of the adder.

Abstract: In PCM systems, it is known to convert .mu.-law digital code words including a sign bit, m characteristic bits, and n mantissa bits representing respectively the polarity, segment value, and quantizing step of a quantized analog sample into a binary floating point representation thereof including a sign, exponent, and mantissa in order to process the digital signal. Known converters are of two relatively expensive types: (1) using a memory having 2.sup.m+n locations or (2) employing a .mu.-law to fixed point conversion followed by fixed point to floating point conversion. To mitigate such drawbacks, the present converter (100) employs a relatively inexpensive translator (200) for translating the segment value and a prefixed quantizing step into a reference mantissa. The code word quantizing step is then added (90, 95) to the reference mantissa to provide the floating point mantissa. The exponent is equal to the segment value; and the sign bit represents the sign.

Abstract: The subject echo canceller includes a controller, which uses a feedback arrangement to improve speech detector performance, for providing an improved multistate control signal responsive to which adjustment of the estimate is inhibited or not inhibited. For example, a first state of the control signal is provided when the value of a first function of the error signal exceeds the value of a second function of the error signal, the second function being an average of first function values. One first function may be the algebraic square of the error signal as normalized with respect to the energy detected in a far end signal, while the second function is the average of the normalized signal where the averaging is done over a predetermined time interval. The predetermined time intervals may be those time intervals during which near end speech is absent.