Electrical audio signal processing systems and devices – Including amplitude or volume control – Automatic
Reexamination Certificate
1997-12-31
2001-10-02
Mei, Xu (Department: 2644)
Electrical audio signal processing systems and devices
Including amplitude or volume control
Automatic
C381S104000, C381S102000
Reexamination Certificate
active
06298139
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to controlling the amount of gain applied to an audio speech signal, and in particular, to maintaining a constant speech envelop by control of the gain.
2. Background of the Art
When processing audio samples, it is often desirable to have the samples at a fixed amplitude. The reason for this is that many such audio processing methods or algorithms are typically optimized for a given level, as is well known in the art. For example, too little gain runs the risk that noise will disturb accurate processing results. Too much gain risks over creating too big a signal. Clipping or other detrimental results can occur.
When processing speech, for example, there is normally a stable level. But in radio communications, for example, if speech is transmitted and the volume of the receiver is turned up, then there can be problems that could disrupt audio processing of the received signal. Therefore, it is many times desirable to keep a constant level for as much of the time speech is present and is to be processed.
Automatic gain control (AGC) has long been used as a method of controlling gain. Some sort of feedback is utilized to compare the current level of a signal with prior levels, or the current levels could be compared to some predetermined level. The circuitry would automatically adjust gain based on that type of monitoring.
AGC methods have been applied to audio envelopes, including the attempt to maintain a fixed level. For example, some current AGC methods are able to adjust the audio to a fixed level by using a rapidly changing gain control. The gain adjusts rapidly whenever the input envelope changes regardless of whether the envelope is increasing or decreasing. These methods work well for data input where the envelope is fairly constant. However, speech tends to have many gaps ranging from very small (those between syllables) to large (those between sentences). Thus, the envelope changes often. Thus, methods exist to detect if a level that is rising is probably speech (as opposed to non-speech or noise). However, there are not good methods to predict if a decrease in level is still speech or whether speech has ended. If the gain is improvidently raised or lowered, then the risk is that background noise will be counter-productive to audio processing. The need for a constant level exists so that the audio processor has the best information about the actual speech or audio.
Therefore, the state of the art attempts to use AGC to rapidly adjust gain up or down depending on the input signal. Consequently, the speech envelope can change up or down by the speech envelope changing over the course of significant speech content (i.e. during the speaking of a sentence). However, using known AGC methods, gain would quickly adjust upwardly when the sentence is started and maintain a constant speech envelope during a syllable. But at the end of a syllable it would likewise start to adjust quickly downward. At the beginning of the next syllable it would adjust quickly up and so on. The speech envelope would likewise increase and decrease rapidly and may not be very constant. Therefore, the goal of the constant speech envelope would be difficult to maintain. The conventional prior art methods determine the level of the audio from a recent sample of audio and compare the recent sample level with a mean of prior sampled levels. Any adjustment in gain is weighted towards a mean value of past samples of the audio signal to avoid quick, directly corresponding changes in gain relative to the signal. For example, some existing methods weight any adjustment in gain 0.9 towards a mean of past samples of the audio versus 0.1 towards a recent sample. This weighting is the same, however, whether gain is adjusted when the signal level is determined to be increasing as well as when it is determined to be decreasing.
Because of the benefits of a constant speech envelope for certain audio processing methods, there is a real need in the art to improve upon the state of the art with regard to the use of AGC methods with audio sample to maintain a fixed speech envelope.
Some prior art AGC methods specify the amount of gain increase or decrease that should take place during AGC. The absolute value of the input signal level is obtained. A mean of prior actual samples is also maintained. Then, the current actual sample is compared to the mean. It is assumed that if the current sample is greater than the mean, that the signal amplitude is increasing. Conversely, if the current sample is less than the mean, it is assumed that the signal level is decreasing.
A pair of coefficients are predetermined. One coefficient essentially controls the amount of decrease of gain applied to the current audio sample if the level of the audio is increasing. The other coefficient determines the amount of increase of gain applied to the current sample if the level of the signal is decreasing. Thus, a single pair of coefficients are fixed over the processing of the speech and generally result in an increase or decrease in gain of the same amount or rate.
It is therefore a principle object of the present invention to provide an apparatus and method for maintaining a constant speech envelope using automatic gain control which improves over or solves the problems of deficiencies in the art.
Further objects of the present invention are to provide an apparatus and method as previously described which:
A. differs with respect to the amount of gain applied when an input signal is increasing versus the amount or the gain applied when the input signal is decreasing.
B. weights the amount of gain applied to an increasing signal more favorably to the level of the current input to achieve a relatively slow and beneficial upward adjustment when the signal is increasing.
C. weights more heavily the applied gain relative to a mean of past input signals when the input signal is decreasing than if the signal were increasing to achieve a slower and thus more beneficial downward adjustment when the signal is decreasing.
D. maintains a relatively constant speech envelope during speech.
E. effectively maintains a fairly constant gain even in between syllables or words of speech.
F. more quickly raises gain when speech occurs after a relatively long period of no speech, but more slowly decreases gain even if it is indicated that the signal level is falling so that gaps between syllables or sentences in speech are not handled as if speech has stopped.
G. is accurate, relatively economical, and durable.
These and other objects, features, and advantages of the present invention will become more apparent with reference to the accompanying specification and claims.
SUMMARY OF THE INVENTION
The present invention comprises an apparatus and method for maintaining a constant speech envelope when using automatic gain control.
The method includes monitoring the actual audio signal amplitude. If increasing, a gain increase of a first amount is applied to the gain signal. The gain signal is the inverse of the mean of prior samples of the signal. If decreasing, a gain of a second amount is applied to the audio signal. The first amount is greater than the second amount of gain. In one option, the invention weights the gain increase of the first amount more heavily towards the level of the most recent part of the audio signal than prior art methods, but on the other hand weights the gain decrease of the second level more heavily towards a mean of past actual audio levels. By doing so, the method allows increase in gain upon an increasing audio signal to bring the speech envelope up to a constant level at the beginning of a portion of speech, but on the other hand causes slower decreases in gain if the level of speech decreases. The method thus creates a more constant speech envelope during speech, which sometimes can include short gaps between syllables and words, and avoids a more non-constant speech envelope if gain were allowed to increase and decrease rapidly with increasing and decre
Poulsen Steven P.
Preston, II James P.
Mei Xu
Transcrypt International, Inc.
Zarley, McKee, Thomte, Voorhes & Sease
LandOfFree
Apparatus and method for maintaining a constant speech... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Apparatus and method for maintaining a constant speech..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Apparatus and method for maintaining a constant speech... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2602696