Adaptive quantizer with instantaneous error robustness

Claim 2

An adaptive quantizer of the type receiving input signal samples and quantizing said samples in accordance with a plurality of quantizing steps, and wherein said adaptive quantizer includes means for varying the size of said quantizing steps in accordance with an adaptation algorithm, the improvement characterized in that said means for varying said quantizing step sizes comprises varying means for varying said quantizing step sizes in accordance with a plurality of preceding quantizer outputs and wherein said varying means provides at least a small leakage in its step size variation at all times and a larger leakage in its step size variation in response to the occurrence of an error in the quantizer output.

Claim 3

In an Adaptive Differential Pulse Code Modulation (ADPCM) encoder of the type wherein a sequence of input signal samples are received and compared in comparison means with at least one predicted value to obtain a difference signal, a quantizer receives and quantizes said difference signal according to a plurality of quantization levels and feedback means generates said at least one predicted value in accordance with the values of a series of said difference signals, the improvement characterized in that: said feedback means comprises first prediction means for generating a first prediction value in accordance with a series of values of said difference signal and for generating a first intermediate signal, and second prediction means for generating a second prediction value in accordance with a series of said first intermediate signals; and said quantizer comprises an adaptive quantizer varying its quantizing steps in accordance with a plurality of preceding quantizer outputs, said adaptive quantizer varying said quantizing step sizes in accordance with an adaptation algorithm and providing at least a small leakage in its step size variation at all times and a larger leakage in its step size variation in response to the occurrence of an error in said quantizer output.

Claim 4

In an Adaptive Differential Pulse Code Modulation (ADPCM) encoder of the type wherein a sequence of input signal samples are received and compared in comparison means with at least one predicted value to obtain a difference signal, a quantizer receives and quantizes said difference signal according to a plurality of quantization levels and feedback means generates said at least one predicted value in accordance with the values of a series of said difference signals, the improvement characterized in that: said feedback means comprises first prediction means for generating a first prediction value in accordance with a series of values of said difference signal and for generating a first intermediate signal, and second prediction means for generating a second prediction value in accordance with a series of said first intermediate signals; and said quantizer comprises an adaptive quantizer varying its quantizing steps in accordance with a plurality of preceding quantizer outputs, said adaptive quantizer varying said quantizing step sizes in accordance with an adaptation algorithm and providing at least a small leakage in its step size variation at all times and a larger leakage in its step size variation in response to the occurrence of an error in said quantizer output.

Claim 5

An adaptive quantizer as defined in claim 1, wherein said adaptation algorithm for defining the quantization step size is given by ##EQU13## where Δ j+1 is the current quantizing step size, Δ j is the previous quantizing step size, (l-α) is a leakage coefficient, M(I j ) is a step size multiplier which is a function of the level I j of the quantizer output, K1 and K2 are predetermined integers and C ij (i=1, 2) are given by ##EQU14## where γi are predetermined parameters and Th i are threshold levels.

Claim 6

In an Adaptive Differential Pulse Code Modulation (ADPCM) encoder of the type wherein a sequence of input signal samples are received and compared in comparison means with at least one predicted value to obtain a difference signal, a quantizer receives and quantizes said difference signal according to a plurality of quantization levels and feedback means generates said at least one predicted value in accordance with the values of a series of said difference signals, the improvement characterized in that said feedback means comprises: first adaptive prediction means for generating a first prediction value from a series of values of said difference signal in accordance with coefficients defined according to a first adaptation algorithm, and for generating a first intermediate signal; and second prediction means for generating a second prediction value from a series of said first intermediate signals in accordance with coefficients defined according to a second adaptation algorithm different from said first adaptation algorithm.

Claim 7

In an Adaptive Differential Pulse Code Modulation (ADPCM) encoder of the type wherein a sequence of input signal samples are received and compared in comparison means with at least one predicted value to obtain a difference signal, a quantizer receives and quantizes said difference signal according to a plurality of quantization levels and feedback means generates said at least one predicted value in accordance with the values of a series of said difference signals, the improvement characterized in that said feedback means comprises: first adaptive prediction means for generating a first prediction value from a series of values of said difference signal in accordance with coefficients defined according to a first adaptation algorithm, and for generating a first intermediate signal; and second prediction means for generating a second prediction value from a series of said first intermediate signals in accordance with coefficients defined according to a second adaptation algorithm different from said first adaptation algorithm.

Claim 8

An encoder as defined in claim 4, wherein said first prediction means comprises: a first adder for receiving as a first input signal a signal representing said difference signal and for receiving a second input signal, the output of said adder comprising said first intermediate signal; and a first predictor for receiving said first intermediate value and generating said first prediction value in accordance with a sequence of said first intermediate values, the output of said first predictor being provided as the second input to said first adder.

Claim 9

An encoder as defined in claim 4, wherein said second prediction means comprises: a second adder for receiving as a first input said intermediate signal and for receiving a second input, the output of said second adder comprising a decoded input signal sample; and a second predictor for receiving said decoded input signal sample and generating said second prediction value in accordance with a sequence of said decoded input signal samples, the output of said second predictor being provided as said second input to said second adder.

Claim 10

An encoder as defined in any one of claims 4-6, wherein said comparison means comprises a first subtractor for subtracting said second predicted value from an input signal sample and a second subtractor for subtracting said first prediction value from the output of said first subtractor in order to obtain said difference signal.

Claim 11

An encoder as defined in claim 5, wherein said first prediction value R ij given by ##EQU15## where a 1j (i) is the coefficient of an i-th tap at a j-th sampling index and S 1j is said first intermediate signal.

Claim 12

An encoder as defined in claim 5, wherein said second prediction value is defined by ##EQU16## where a 2j (i) is the coefficient of an i-th tap at a j-th sampling index, and S j is said locally decoded input signal sample.

Claim 13

An encoder as defined in claim 8, wherein said first predictor is an adaptive predictor whose coefficients are updated by: ##EQU17## where (l-βl) represents a leakage factor, g 1 is a predetermined convergence rate parameter, V j is a quantizer output level with unit variance, E j is a signal representing said error signal value and ξ i and η i are parameters defining optimal step size for zero mean and unit variance of the distribution of input signal samples.

Claim 14

An encoder as defined in claim 10, wherein said second predictor is an adaptive predictor whose coefficients are updated by: ##EQU18## where g 2 is a predetermined convergence rate parameter.

Claim 15

An encoder as defined in claim 5, further comprising an inverse quantizer for receiving the output of said quantizer and providing said difference signal value to said first adder.

Claim 16

An encoder as defined in claim 4, wherein said quantizer is an adaptive quantizer varying its quantizing step sizes in accordance with a plurality of preceding quantizer outputs.

Claim 17

An encoder as defined in claim 13, wherein said adaptive quantizer varies said step sizes in accordance with an adaptation algorithm and provides at least a small leakage in its step size variation at all times and a larger leakage in its step size variation in response to the occurrence of an error in said quantizer output.