Digital automatic gain control with lookahead, adaptive noise floor sensing, and decay boost initialization

Claim 3

The method of claim 1, wherein the steps for initialization are modified by an initial gain floor level estimate stored in the memory and further comprise the steps of detecting valid speech in the signal sample by sensing a signal level of the current subframe, adjusting the floor level estimate in response to the signal level of the current subframe, and finding valid speech when a plurality of consecutive subframes each have a signal level greater than the sum of the floor level estimate and an interval value.

Claim 4

The method of claim 2, wherein the steps for detecting valid speech further comprise a step in response to absence of valid speech of adjusting a noise floor threshold in the memory when the sum of the total signal level of a plurality of subframes and an adjust factor is less than the noise floor threshold.

Claim 5

The method of claim 3, wherein the system further comprises a table in the memory containing pairs of level means entries and gain entries and wherein the steps of causing gain attack further comprise the step of selecting one of the gain entries from the table as an attack level when the total signal level matches one of the level mean entries corresponding to the selected gain entry.

Claim 6

An automatic gain controller for a digitized audio signal, comprising: a memory comprising a buffer, the buffer comprising a plurality of subframes, each subframe containing digitized data samples of the signal, the subframes including at least one future subframe and a current subframe; and signal processing means coupled to the memory for controlling gain of the audio signal represented by the current subframe in the buffer comprising means for computing a plurality of mean signal level values from the plurality of subframes, each mean signal level value in the plurality corresponding to one of the subframes; means for causing decay of gain on the signal represented by the current subframe when a first set of the mean values are each below a low threshold signal level; means for causing gain attack on the audio signal represented by the current subframe when a second set of the mean values are each above a high threshold signal level; means for shifting the data samples in at least one future subframe into the current subframe after causing gain attack or decay for controlling gain of the audio signal; thereby causing gain attack and decay on the audio signal represented by the subframes which attack and decay appear to occur without delay to a user of the audio signal; and, means for estimating background noise of the signal sample by setting a background noise level value in the memory and adjusting the noise level value to match the lowest of the mean signal level values.

Claim 7

An automatic gain controller for a digitized audio signal, comprising: a memory comprising a buffer, the buffer comprising a plurality of subframes, each subframe containing digitized data samples of the signal, the subframes including at least one future subframe and a current subframe; and signal processing means coupled to the memory for controlling gain of the audio signal represented by the current subframe in the buffer comprising means for computing a plurality of mean signal level values from the plurality of subframes, each mean signal level value in the plurality corresponding to one of the subframes; means for causing decay of gain on the signal represented by the current subframe when a first set of the mean values are each below a low threshold signal level; means for causing gain attack on the audio signal represented by the current subframe when a second set of the mean values are each above a high threshold signal level; means for shifting the data samples in at least one future subframe into the current subframe after causing gain attack or decay for controlling gain of the audio signal; thereby causing gain attack and decay on the audio signal represented by the subframes which attack and decay appear to occur without delay to a user of the audio signal; and, means for estimating background noise of the signal sample by setting a background noise level value in the memory and adjusting the noise level value to match the lowest of the mean signal level values.

Claim 8

An automatic gain controller for a digitized audio signal, comprising: a memory comprising a buffer, the buffer comprising a plurality of subframes, each subframe containing digitized data samples of the signal, the subframes including at least one future subframe and a current subframe; and signal processing means coupled to the memory for controlling gain of the audio signal represented by the current subframe in the buffer comprising means for computing a plurality of mean signal level values from the plurality of subframes, each mean signal level value in the plurality corresponding to one of the subframes; means for causing decay of gain on the signal represented by the current subframe when a first set of the mean values are each below a low threshold signal level; means for causing gain attack on the audio signal represented by the current subframe when a second set of the mean values are each above a high threshold signal level; means for shifting the data samples in at least one future subframe into the current subframe after causing gain attack or decay for controlling gain of the audio signal; thereby causing gain attack and decay on the audio signal represented by the subframes which attack and decay appear to occur without delay to a user of the audio signal; initialization means coupled to the signal processing means for setting an initial signal gain level for the controller and having an attack level setting in the memory; a decay level setting in the memory; a desired gain level setting in the memory; means for increasing the attack level setting by a controlled amount during each of the predetermined initial plurality of subframes processed by the controller; and means for disabling the initialization means when the attack level setting equals or exceeds the desired gain level; and means for estimating background noise of the signal sample by setting a background noise level value in the memory and adjusting the noise level value to match the lowest of the mean signal level values.

Claim 9

An automatic gain controller for a digitized audio signal, comprising: a memory comprising a buffer, the buffer comprising a plurality of subframes, each subframe containing digitized data samples of the signal, the subframes including at least one future subframe and a current subframe; and signal processing means coupled to the memory for controlling gain of the audio signal represented by the current subframe in the buffer comprising means for computing a plurality of mean signal level values from the plurality of subframes, each mean signal level value in the plurality corresponding to one of the subframes; means for causing decay of gain on the signal represented by the current subframe when a first set of the mean values are each below a low threshold signal level; means for causing gain attack on the audio signal represented by the current subframe when a second set of the mean values are each above a high threshold signal level; means for shifting the data samples in at least one future subframe into the current subframe after causing gain attack or decay for controlling gain of the audio signal; thereby causing gain attack and decay on the audio signal represented by the subframes which attack and decay appear to occur without delay to a user of the audio signal; initialization means coupled to the signal processing means for setting an initial signal gain level for the controller and having an attack level setting in the memory; a decay level setting in the memory; a desired gain level setting in the memory; means for increasing the attack level setting by a controlled amount during each of the predetermined initial plurality of subframes processed by the controller; and means for disabling the initialization means when the attack level setting equals or exceeds the desired gain level; and means for estimating background noise of the signal sample by setting a background noise level value in the memory and adjusting the noise level value to match the lowest of the mean signal level values.

Claim 10

The controller of claim 6, wherein the initialization means further comprises: a gain floor level value stored in the memory; means for detecting valid speech in the signal sample, the detecting means comprising means for: sensing an input signal level represented by the current subframe; adjusting the floor level value in response to the input signal level represented by the current subframe; and finding valid speech when a plurality of consecutive subframes each contain signal data having a signal level greater than the sum of the floor level value and an interval level.

Claim 11

The controller of claim 6, wherein the signal processing means further comprising means for sensing presence of a valid speech signal data sample in the subframes and dependent thereon, selectively causing gain decay.

Claim 12

The controller of claim 7, wherein the means for sensing valid speech further comprises means responsive to absence of a valid speech data sample for adjusting the floor level value in the memory when the mean signal level value in a plurality of subframes and an adjust factor is less than the floor level value.

Claim 13

An automatic gain controller for a digitized audio signal, comprising: a memory comprising a buffer, the buffer comprising a plurality of subframes, each subframe containing digitized data samples of the signal, the subframes including at least one future subframe and a current subframe; and signal processing means coupled to the memory for controlling gain of the audio signal represented by the current subframe in the buffer comprising means for computing a plurality of mean signal level values from the plurality of subframes, each mean signal level value in the plurality corresponding to one of the subframes; means for causing decay of gain on the signal represented by the current subframe when a first set of the mean values are each below a low threshold signal level; means for causing gain attack on the audio signal represented by the current subframe when a second set of the mean values are each above a high threshold signal level, said second set comprising a number of subframes determined by a variable spike resistance parameter in the memory; and means for shifting the data samples in at least one future subframe into the current subframe after causing gain attack or decay for controlling gain of the audio signal; thereby causing gain attack and decay on the audio signal represented by the subframes which attack and decay appear to occur without delay to a user of the audio signal.

Claim 14

An automatic gain controller for a digitized audio signal, comprising: a memory comprising a buffer, the buffer comprising a plurality of subframes, each subframe containing digitized data samples of the signal, the subframes including at least one future subframe and a current subframe; and signal processing means coupled to the memory for controlling gain of the audio signal represented by the current subframe in the buffer comprising means for computing a plurality of mean signal level values from the plurality of subframes, each mean signal level value in the plurality corresponding to one of the subframes; means for causing decay of gain on the signal represented by the current subframe when a first set of the mean values are each below a low threshold signal level; means for causing gain attack on the audio signal represented by the current subframe when a second set of the mean values are each above a high threshold signal level, said second set comprising a number of subframes determined by a variable spike resistance parameter in the memory; and means for shifting the data samples in at least one future subframe into the current subframe after causing gain attack or decay for controlling gain of the audio signal; thereby causing gain attack and decay on the audio signal represented by the subframes which attack and decay appear to occur without delay to a user of the audio signal.

Claim 15

An automatic gain controller for a voice processing system, the system having a memory, a mass storage device, and at least one digital signal processor coupled to the memory and the mass storage device, the controller comprising: a buffer in the memory comprising a plurality of subframes organized as a group of future subframes and a current subframe, each subframe containing a digitized signal sample; and data processing means coupled to the memory for recording the samples on the mass storage device after adjusting the gain of the samples, comprising means for computing a mean signal level for each sample in all the subframes; means for causing gain decay on the signal represented by the current subframe when a signal level of samples in a first set of the subframes is below a low threshold signal level; means for causing gain attack on the signal represented by the current subframe when a signal level of samples in a second set of the subframes is above a high threshold signal level; means for shifting the data sample in the current subframe to an output for recording on the mass storage device and for shifting data of one subframe in the future group into the current subframe; and means for estimating background noise of the signal sample by setting a background noise level value in the memory and adjusting the noise level value in response to the lowest of the mean signal level values.

Claim 16

An automatic gain controller for a voice processing system, the system having a memory, a mass storage device, and at least one digital signal processor coupled to the memory and the mass storage device, the controller comprising: a buffer in the memory comprising a plurality of subframes organized as a group of future subframes and a current subframe, each subframe containing a digitized signal sample; and data processing means coupled to the memory for recording the samples on the mass storage device after adjusting the gain of the samples, comprising means for computing a mean signal level for each sample in all the subframes; means for causing gain decay on the signal represented by the current subframe when a signal level of samples in a first set of the subframes is below a low threshold signal level; means for causing gain attack on the signal represented by the current subframe when a signal level of samples in a second set of the subframes is above a high threshold signal level; means for shifting the data sample in the current subframe to an output for recording on the mass storage device and for shifting data of one subframe in the future group into the current subframe; and means for estimating background noise of the signal sample by setting a background noise level value in the memory and adjusting the noise level value in response to the lowest of the mean signal level values.

Claim 17

An automatic gain controller for a voice processing system, the system having a memory, a mass storage device, and at least one digital signal processor coupled to the memory and the mass storage device, the controller comprising: a buffer in the memory comprising a plurality of subframes organized as a group of future subframes and a current subframe, each subframe containing a digitized signal sample; and data processing means coupled to the memory for recording the samples on the mass storage device after adjusting the gain of the samples, comprising means for computing a mean signal level for each sample in all the subframes; means for causing gain decay on the signal represented by the current subframe when a signal level of samples in a first set of the subframes is below a low threshold signal level; means for causing gain attack on the signal represented by the current subframe when a signal level of samples in a second set of the subframes is above a high threshold signal level; means for shifting the data sample in the current subframe to an output for recording on the mass storage device and for shifting data of one subframe in the future group into the current subframe; initialization means coupled to the digital signal processing means for setting an initial gain level of the controller and having an attack level setting in the memory; a decay level setting in the memory; a desired gain level setting in the memory; means for increasing the attack level setting by a controlled amount during processing of each of a predetermined plurality of frames processed by the controller at the start of a frame stream; means for disabling the initialization means when the attack level setting equals or exceeds the desired gain level; and means for estimating background noise of the signal sample by setting a background noise level value in the memory and adjusting the noise level value in response to the lowest of the mean signal level value.

Claim 18

An automatic gain controller for a voice processing system, the system having a memory, a mass storage device, and at least one digital signal processor coupled to the memory and the mass storage device, the controller comprising: a buffer in the memory comprising a plurality of subframes organized as a group of future subframes and a current subframe, each subframe containing a digitized signal sample; and data processing means coupled to the memory for recording the samples on the mass storage device after adjusting the gain of the samples, comprising means for computing a mean signal level for each sample in all the subframes; means for causing gain decay on the signal represented by the current subframe when a signal level of samples in a first set of the subframes is below a low threshold signal level; means for causing gain attack on the signal represented by the current subframe when a signal level of samples in a second set of the subframes is above a high threshold signal level; means for shifting the data sample in the current subframe to an output for recording on the mass storage device and for shifting data of one subframe in the future group into the current subframe; initialization means coupled to the digital signal processing means for setting an initial gain level of the controller and having an attack level setting in the memory; a decay level setting in the memory; a desired gain level setting in the memory; means for increasing the attack level setting by a controlled amount during processing of each of a predetermined plurality of frames processed by the controller at the start of a frame stream; means for disabling the initialization means when the attack level setting equals or exceeds the desired gain level; and means for estimating background noise of the signal sample by setting a background noise level value in the memory and adjusting the noise level value in response to the lowest of the mean signal level value.

Claim 19

The controller of claim 11 or 12, wherein the initialization means for further comprises: an initial gain floor level estimate stored in the memory; means for detecting valid speech in the signal sample comprising means for: sensing an input signal level of the current subframe; adjusting the floor level estimate in response to the input signal level of the current subframe; and finding valid speech when a mean signal level of a plurality of consecutive subframes is greater than the sum of the floor level estimate and an interval level.

Claim 20

The controller of claim 13, wherein the means for detecting valid speech further comprises means for finding valid speech when a plurality of consecutive subframes each have a signal level above the floor level estimate.

Claim 21

The controller of claim 14, wherein the means for sensing valid speech further comprises means responsive to absence of valid speech for adjusting the floor level estimate when the sum of the total signal level of a plurality of subframes and an adjust factor is less than the floor level estimate.

Claim 22

The controller of claim 15 wherein the means for causing gain attack further comprises: a table in the memory containing pairs of mean level entries and gain entries; and means for selecting one of the gain entries from the table as an attack level when the total signal level matches one of the level mean entries corresponding to the selected gain entry.

Claim 23

An automatic gain controller for a voice processing system, the system having a memory, a mass storage device, and at least one digital signal processor coupled to the memory and the mass storage device, the controller comprising: a buffer in the memory comprising a plurality of subframes organized as a group of future subframes and a current subframe, each subframe containing a digitized signal sample; and data processing means coupled to the memory for recording the samples on the mass storage device after adjusting the gain of the samples, comprising means for computing a mean signal level for each sample in all the subframes; means for causing gain decay on the signal represented by the current subframe when a signal level of samples in a first set of the subframes is below a low threshold signal level; means for causing gain attack on the signal represented by the current subframe when a signal level of samples in a second set of the subframes is above a high threshold signal level said second set comprising a number of subframes determined by a variable spike resistance parameter in the memory; and means for shifting the data sample in the current subframe to an output for recording on the mass storage device and for shifting data of one subframe in the future group into the current subframe.

Claim 24

An automatic gain controller for a voice processing system, the system having a memory, a mass storage device, and at least one digital signal processor coupled to the memory and the mass storage device, the controller comprising: a buffer in the memory comprising a plurality of subframes organized as a group of future subframes and a current subframe, each subframe containing a digitized signal sample; and data processing means coupled to the memory for recording the samples on the mass storage device after adjusting the gain of the samples, comprising means for computing a mean signal level for each sample in all the subframes; means for causing gain decay on the signal represented by the current subframe when a signal level of samples in a first set of the subframes is below a low threshold signal level; means for causing gain attack on the signal represented by the current subframe when a signal level of samples in a second set of the subframes is above a high threshold signal level said second set comprising a number of subframes determined by a variable spike resistance parameter in the memory; and means for shifting the data sample in the current subframe to an output for recording on the mass storage device and for shifting data of one subframe in the future group into the current subframe.

Claim 31

The method of claim 19, wherein the initialization steps further comprise steps for detecting valid speech in the signal sample by sensing an input signal level of the current subframe, adjusting the noise level value in response to the input signal level of the current subframe, and finding valid speech when a plurality of consecutive subframes each have a signal level greater than the sum of the noise level value and an interval level.

Claim 32

The method of claim 21, further comprising steps for detecting the presence of a valid speech data sample in the subframes when a plurality of consecutive subframes each have a signal level above an absolute threshold value stored in the memory.

Claim 33

The method of claim 22, wherein the step of detecting valid speech further comprises steps for responding to absence of a valid speech data sample by for adjusting the floor threshold value in the memory when the mean signal level value in a plurality of subframes and an adjust factor is less than the floor threshold value.

Claim 34

The method of claim 23, wherein the memory contains a table containing pairs of corresponding level mean entries and gain entries, and wherein the step of causing gain attack further comprises the substep of selecting one of the gain entries from the table when the total signal level matches one of the level mean entries corresponding to the selected gain entry, and controlling the step of causing gain attack in accordance with the selected gain entry.