GB2407012A

GB2407012A - Speech communication unit and method of level adjustment therein

Info

Publication number: GB2407012A
Application number: GB0322734A
Authority: GB
Inventors: Uzi Cohen; Itzhak Avayu; Boaz Or-Shraga
Original assignee: Motorola Inc
Current assignee: Motorola Solutions Inc
Priority date: 2003-09-27
Filing date: 2003-09-27
Publication date: 2005-04-13
Anticipated expiration: 2023-09-27
Also published as: GB0322734D0; GB2407012B

Abstract

A speech communication unit (200) comprises a transmitter for transmitting speech signals and/or a receiver for receiving speech signals. The transmitter and/or receiver comprises a speech processor (209) and a signal level adjustment circuit (226), operably coupled to the speech processor (209), for adjusting an amplitude level of a speech signal. The speech processor (209) calculates a number of clipped speech samples in a first period of time; and adjusts a gain level of the signal level adjustment circuit (226) in a second period of time, in response to the number of clipped speech samples. A method of adjusting a level of a speech signal in a speech communication unit is also described. In this manner, a speech communication unit is able to track rapidly changing saturated speech signals, and adjust gain/attenuator levels in a receiver or transmitter chain in response thereto.

Description

24070 1 2 - 1 -

SPEECH COMMUNICATION UNIT AND METHOD OF LEVEL ADJUSTMENT

THEREIN

Field of the Invention

This invention relates to signal estimation technology in a speech communication unit. The invention is applicable to, but not limited to, a gain control mechanism to reduce distortion due to clipping of speech signals.

Background of the Invention

In the field of this invention, it is known that most wireless communication units apply some form of automatic gain control (AGC), to automatically adjust a gain/attenuation applied to received or transmitted signals. AGC is, in effect, a processor controlled mechanism to control the gain applied to a signal throughout any or all stages of the transmitter or receiver circuitry. Such AGC operation needs to be wideband in nature when some portion of the AGC circuitry is operational at radio frequencies (RF), or may be narrow- band in nature if the AGC circuitry is only operational at intermediate (IF) or baseband frequencies.

AGC circuits and associated algorithms are often used in wireless speech communication units in order to free a user from constantly making audio volume adjustments. A generic block diagram 100 of an AGC circuit is shown in FIG. 1. An input analogue signal 105 is amplified in amplifier 110 by a predetermined gain and then digitised using the analogue to digital A/D converter 115. The output samples 120 are then processed over a long period of time in a gain estimator block 125 to determine the - 2 signal's amplitude level. The gain estimator block 125 is then able to determine a level that should be applied to attenuate or amplify the speech signal to yield a nominal' signal level. The gain estimator block 125 successively adjusts the gain value applied by the analogue amplifier 110 until the signal level output from the A/D converter 115 is acceptable.

In this case, an AGC algorithm will attenuate a signal gain applied to a signal level when a talker's utterances are so loud that there is a nonlinear output from the A/D converter 115, i.e. the speech signal is saturated.

The basic operation of the AGC is to adjust the input signal to a predetermined nominal level.

When a sudden burst of speech occurs, it is possible that the input signal may be amplified to a level where it is clipped. For such a case, the recovered (synthesized) voice signal will sound distorted, due to voice pumping (increase and decrease of gain level) that degrades the perceived voice quality.

However, in the context of AGC circuits for speech communication, there is no current known way to estimate quickly an amount of attenuation needed to be applied to a speech signal. In this regard, a convergence feedback mechanism takes too much time to adjust a bursty signal level until the appropriate level of attenuation or amplification is applied.

In summary, it is known that the provision of stable, AGC circuits in wireless communication units is complex.

Furthermore, the known AGC 'convergence' mechanisms are too slow to adequately handle bursty speech - 3 communication. Thus, there exists a need for an improved AGC circuit and corresponding AGC algorithm to provide a faster convergence for speech communication, wherein the abovementioned disadvantages may be alleviated. s

Statement of Invention

In accordance with a first aspect of the present invention, there is provided a speech communication unit.

The speech communication unit comprises a transmitter for transmitting speech signals and/or a receiver for receiving speech signals. The transmitter and/or receiver comprise a speech processor and a signal level adjustment circuit, operably coupled to the speech processor, for adjusting an amplitude level of a speech signal. The speech processor calculates a number of clipped speech samples in a first period of time; and adjusts a gain level of the signal level adjustment circuit in a second period of time, in response to the number of clipped speech samples.

In accordance with a second aspect of the present invention, there is provided a method of adjusting a level of a speech signal in a speech communication unit.

The method comprises the steps of receiving or transmitting a speech signal; processing the speech signal; and adjusting a gain applied to the speech signal based on the processed speech signal. The method further comprises counting a number of clipped speech samples in a first period of time such that the step of adjusting a gain is performed in a second period of time, in response to the number of clipped speech samples.

Further features of the present invention are as claimed in the dependent Claims.

Thus, the speech processing function calculates (or estimates) a degree of speech signal saturation by calculating or estimating a number of clipped speech samples. As the calculation is made substantially in a real-time manner, it is possible for the speech processing function to rapidly adjust the gain (or attenuation) to be applied to the speech signal to reduce (or even prevent) noise distortion due to clipping of a speech signal. Preferably, the periods of time are speech frames; so that the attenuation/gain applied to consecutive speech frames can be continuously optimised.

Brief Description of the Drawings

FIG. l shows a simplified block diagram of a known AGC circuit.

Exemplary embodiments of the present invention will now be described, with reference to the accompanying drawings, in which: FIG. 2 illustrates a block diagram representation of a wireless communication unit adapted to employ automatic gain control for a speech signal in accordance with a preferred embodiment of the present invention; FIG. 3 illustrates a graph showing an example of an amplitude distribution of a speech signal; FIG. 4 illustrates a flowchart of an automatic gain control mechanism to be applied to a speech signal in - 5 - accordance with a preferred embodiment of the present invention; and FIG. 5 illustrates a graph of an average number of clipped samples in a speech frame versus a gain to be applied to a speech signal, in accordance with a preferred embodiment of the present invention.

Description of Preferred Embodiments

The preferred embodiment of the present invention relates to signal level estimation technology for a speech communication unit. The inventive concepts are described below in the context of a transmitter in a wireless communication unit. However, a skilled artisan will appreciate that the inventive concepts can be equally applied in a receiver context, for example when a radio frequency signal is detected and digitized within the wireless communication unit.

Referring now to FIG. 2, a block diagram of a wireless subscriber unit/mobile station (MS) 200, adapted to support the inventive concepts of the preferred embodiments of the present invention, is shown. The MS 200 contains an antenna 202 preferably coupled to a duplex filter or antenna switch 204 that provides isolation between receive and transmit chains within MS 200. The receiver chain includes receiver front-end circuitry 206 (effectively providing reception, filtering and intermediate or base-band frequency conversion). The front-end circuit 206 is serially coupled to a signal processing function (generally realised by a digital signal processor (DSP)) 208 via a baseband (back-end) processing circuit 207. A controller 214 is operably - 6 - coupled to the front-end circuitry 206 and a received signal strength indication (RSSI) function 212; so that the receiver can calculate receive bit-error-rate (BER) or frame-error-rate (PER) or similar link- quality measurement data from recovered information via the RSSI function 212.

A memory device 216 is operably coupled to the signal processing function 208 and controller 214 and stores a wide array of MS-specific data, for example decoding/encoding information. A timer 218 is operably coupled to the controller 214 to control the timing of operations, namely the transmission or reception of time- dependent signals such as the speech frames, within the MS 200. As known in the art, received signals that are processed by the signal processing function 208 are typically applied to an output device 210, such as a speaker.

The transmit chain comprises a speech input device 220, such as a microphone, coupled in series through a baseband processing circuit 222 to the processor 208. An output from the signal processing function 208 is input to the transmitter/modulation circuitry 228 and thereon to a power amplifier 230. The signal processing function 208, transmitter/modulation circuitry 228 and the power amplifier 230 are operationally responsive to the controller 214, with an output from the power amplifier coupled to the duplex filter or antenna switch 204, as known in the art.

In accordance with a preferred embodiment of the present invention, the signal processing function 208 has been adapted to comprise, or be operably coupled to, a speech - 7 - processing function 209, configured to implement the inventive concepts herein described. In particular, the speech processing function 209 has been adapted for a transmitting MS to estimate a number of clipped samples output from the A/D converter 224. In speech signals, the input signal through the microphone is digitized in the A/D converter 224. This digitization can cause clipping'. Clipping is a situation were the signal entering the A/D converter 224 exceeds the functional range of the A/D converter 224, i.e. greater than '+1' or less than '-1'.

The estimate is preferably made in a relatively short period of time, say a speech frame, to determine an appropriate level to adjust a gain applied to the speech signal in a subsequent period of time, say the next speech frame. The adjustment of the speech level is implemented by the speech processing function 209 setting (in a timescale approaching real-time) a gain/attenuation level in a level adjustment circuit 226 contained in the baseband processing circuit 222. The amplitude-adjusted speech, in say the subsequent frame, is then input to the A/D converter 224 at such a level that the A/D converter does not clip the speech signal.

In accordance with the preferred embodiment of the present invention, the memory device 216 has also been adapted to store a look-up table of gain/attenuation settings, or a transform function, to be used by the speech processing function 209 for calculating a level of speech saturation, for example an average number of clipped speech samples in a speech frame. In this manner, the look-up table is preferably accessed by the speech processing function 209 to identify a - 8 - gain/attenuation level to be applied to the level adjustment circuit 226.

It is within the contemplation of the invention that the signal processing function 208 and controller 214 arrangement may be combined into one element, operably connected or separate functions interconnected in a reasonable manner as known in the art. Furthermore, it is envisaged that the signal processing function 208 and/or speech processing function 209 in the transmit chain may be implemented as distinct from the corresponding processor function(s) in the receive chain.

Alternatively, a single processor 208 may be used to implement the processing of both transmit and receive signals, as shown in FIG. 2.

Of course, the various components within the MS 200 can be realised in discrete or integrated component form.

Furthermore, it is within the contemplation of the invention that the MS 200 may be any radio transmitter and/or receiver device, such as a portable or mobile PMR radio, a mobile phone, a personal digital assistant, wireless-capable laptop computer.

In accordance with the preferred embodiment of the present invention, the speech processing function 209 has been adapted to estimate/calculate a level of speech signal saturation within a relatively short period of time, such as a single received speech frame. As an example, the preferred embodiment of the present invention uses a 30 msec. speech frame, with 240 samples at 8 KHz, to provide a good indication of a speech saturation level. A skilled artisan will appreciate that alternative speech periods and sampling rates can be used - 9 - and that in such cases a different transform function may be required.

Advantageously, such a calculation/estimation enables the speech processing function 209 to determine immediately whether the input speech signal is clipped. The only requirement is that a sufficient number of samples are measured to provide a reasonable representation of the speech signal. If the input speech signal is determined as being clipped, the speech processing function 209 is able to immediately attenuate the input speech signal preferably within the next single speech frame. In this manner, the transmitter is able to cope with rapid audio level changes and adjust a gain function to avoid speech clipping by an A/D converter, particularly in a bursty communication channel.

It is envisaged that the speech processing function 209 that calculates a speech saturation level could also be used for distortion indication for various algorithms such as noise-suppression echo-cancelling, vocoders as well as AGC.

In particular, the preferred embodiment of the present invention proposes that the speech processing function 209 estimates/calculates a saturated speech level based on a statistical behaviour of the speech signal's amplitude. In this regard, the speech processing function 209 utilises the fact that a speech signal's amplitude has a Laplace distribution, as shown in FIG. 3.

Referring now to FIG. 3, a graph 300 illustrates a speech sample distribution of the amplitude of a normalised speech signal. Each 'bin' 305 represents an interval of - 10 speech amplitudes, where the height is a number of samples 310 where their amplitude falls in that interval.

For example, the tallest bin represents approximately 40,000 samples that have a value between approximately ' 0.05' to '+0.01'.

In FIG. 3, it can be clearly seen that speech signals are distributed in a fast decaying manner from the centred zero value to '+1'. In a digital domain all speech samples are measured between '+1' and '-1'. When a signal is greater than the maximum A/D boundary, the A/D assigns the value '+1'. Therefore, in order to estimate how many samples were clipped during a sampling period, a count of the number of samples that exceed '+ 0.99' or go below '-0.99' can be performed. In the illustrated case 315, no speech samples had been clipped as there are no bins near the '+ 1' or '-1' values.

Referring now to FIG. 4, a flowchart 400 illustrates the method of adjusting a speech signal, in accordance with the preferred embodiment of the present invention. The method comprises the step of receiving and processing a speech signal in step 402. The preferred method starts with defining two variables in step 405. The variables comprise a 'count' and an 'index' value in order to determine a number of clipped speech samples in the received speech signal. A skilled artisan will appreciate that any of a number of mechanisms can be used to determine how 'saturated' a speech signal is, and that the use of these two variables is only a preferred

example.

The speech processor identifies a first sample value of the speech and compares the value with the clipping - 11 - threshold. The clipping threshold is selected to be a threshold that bounds the desired dynamic range of the speech signal. If the absolute value of the sample exceeds the clipping threshold in step 415, both the S 'count' variable and the 'index' variable are incremented in step 420. If the absolute value of the sample does not exceed the clipping threshold in step 415, the index' variable only is incremented in step 425.

If the sample is not the last sample in the speech frame, i.e. the 'index' variable is less than the number of possible speech samples, in step 430, the process loops back to step 410 to compare the next sample in the speech frame.

Thus, by the time the loop is completed for all the speech frame's samples, the 'count' variable indicates the number of clipped samples within the speech frame, and 'index' is the number of speech samples that have been analysed within a frame.

Once a whole speech frame has been analysed, the next step is to convert the number of clipped samples (i.e. the 'Count' value) into a gain estimation value for setting/adjusting an attenuator or amplifier in the transmitter/receiver chain, as shown in step 435. A skilled artisan will appreciate that the appropriate gain level can be obtained in a number of ways, with the preferred mechanism using a look-up table.

Alternatively, a function that describes a relationship between the variables could be used, or indeed any other appropriate transform, and the inventive concepts are therefore not limited to calculating an 'average' number of speech samples.

Advantageously, the speech processor is then able to adjust a gain element in the transmitter path when receiving the next speech frame from the microphone, to compensate for the level of clipping identified in the previous speech frame.

FIG. 5 shows a graph 500 illustrating an example of speech samples and statistics 515, as collected by the inventors of the present invention. Such a set of sample and statistics can be used in generating the data for the aforementioned look-up table or defining a transform function to be applied to speech signals.

The graph in FIG. 5 is obtained by counting a number of clipped speech samples per frame for saturated audio signals when applying various gains. FIG. 5 indicates a gain level 505 applied to a received speech signal based on an average number of clipped samples in a speech frame 510. The generation of a graph according to FIG. 5, enables a transform (relationship) function to be determined, or a look-up table to be generated for subsequent use in a preferred signal level estimation algorithm.

Thus, by subsequently applying transform function or using the look-up table, a speech processor is able to achieve substantially nominal speech signal levels. In a nominal signal level, there should be no clipping at all, and the whole signal will be substantially confined to the dynamic range of the A/D. This is dependent upon any margin introduced into the optimal gain setting versus a potential rate of speech clipping. - 13

A great advantage of this criterion is the ability to immediately estimate a level of speech saturation for each audio frame analysed. Consequently, rapid changes in a speech signal level can be detected and an immediate gain change applied to overcome clipping and speech distortion.

In the preferred embodiment of the present invention, a clipping count is selected as being frame-based because of the statistical nature of the frame-based speech that constructed the lookup-table. It is envisaged that alternative criteria may be used for other embodiments, so long as a statistically reliable number of samples are used.

In accordance with the preferred embodiment of the present invention, a look-up table may be constructed by applying different gains to a nominal level speech signal. A 'count' may then be used that relates to the number of samples crossing a certain threshold.

Preferably, the measurements are then averaged for various speakers (i.e. male or female). Thus, it is envisaged that a number of look-up tables may be generated for subsequent use, dependent upon the speaker concerned, thereby further improving the flexibility of the present invention.

The inventors of the present invention have determined that the aforementioned technique is noise robust because in most noise scenarios the amplitude statistics of the speech signal does not change. For example, in additive noise cases, the amplitude of the signal is generally higher and causes more clipping. However, when additive noise is applied to the speech processor of the present - 14 invention, it was found that the behaviour is effectively the same as a case where there is no noise.

It is within the contemplation of the invention that alternative transmitter and/or receiver architectures can be used that would also benefit from the inventive concepts described herein, such as direct conversion transmitters/receivers, super heterodyne receivers, etc. In effect, it is envisaged that the inventive concepts hereinbefore described can be applied to any speech communication unit (wireless or wireline) that would benefit from reacting to rapidly changing speech levels in a fast and robust manner, particularly in a noisy environment.

It will be understood that the speech communication unit and method of receiving a speech signal as described above, tend to provide at least one or more of the following advantages: (i) The inventive concepts provide a good estimate of the speech signal distortion, which can be obtained from measuring a number of clipped samples, say in a speech frame; (ii) The inventive concepts enable tracking of rapidly changing saturated speech signals, and thereafter adjustment of gain/attenuator levels in the transmitter and/or receiver chain in response thereto; (iii) The inventive concept is noise robust; and (iv) The inventive concepts are easy to implement on a real-time processor.

Whilst specific, and preferred, implementations of the present invention are described above, it is clear that - 15 one skilled in the art could readily apply variations and modifications of such inventive concepts.

Thus, an improved speech communication unit and S corresponding method for receiving or transmitting a speech signal have been described that aim to provide, for example, a faster response time in implementing an AGO function for speech communication, thereby alleviating at least some of the abovementioned disadvantages with known receiver technology.

Claims

Claims 1. A speech communication unit (200) having a transmitter for

transmitting speech signals and/or a receiver for receiving speech signals, the transmitter and/or receiver comprising: a speech processor (209); and a signal level adjustment circuit (226), operably coupled to the speech processor (209), for adjusting an amplitude level of a speech signal; wherein the transmitter and/or receiver is characterized in that the speech processor (209) calculates a number of clipped speech samples in a first period of time; and adjusts a gain level of the signal level adjustment circuit (226) in a second period of time, in response to the number of clipped speech samples.
2. A speech communication unit (200) according to Claim 1 further characterized in that the speech processor (209) is operably coupled to an analogue to digital converter (224) such that the signal level adjustment circuit (226) adjusts a speech signal level to substantially prevent or reduce an occurrence of the analogue to digital converter (224) clipping the speech signal.
3. A speech communication unit (200) according to Claim 1 or Claim 2 further characterized in that the first and second period of times are speech frames.
4. A speech communication unit (200) according to Claim 3 further characterized in that the first and second period of times are consecutive speech frames or speech frames that are non-time-dispersed. - 17
5. A speech communication unit (200) according to any preceding Claim, wherein the speech communication unit (200) is a wireless speech communication unit and the speech processor controls a gain function in the signal level adjustment circuit when operating in a noisy environment.
6. A speech communication unit (200) according to any preceding Claim further characterized in that a number of clipped speech samples are calculated by counting a number of speech samples that exceed a threshold.
7. A speech communication unit (200) according to any IS preceding Claim, wherein the speech communication unit (200) is one of a portable or mobile radio, a mobile phone, a personal digital assistant, a wireless capable laptop computer.
8. A method of adjusting a level of a speech signal (400) in a speech communication unit comprising the steps of: receiving or transmitting (402) a speech signal; processing (402) the speech signal; and adjusting a gain (435) applied to the speech signal based on the processed speech signal; wherein the method is characterized by the step of: counting a number of clipped speech samples (420) in a first period of time such that said step of adjusting a gain is performed in a second period of time, in response to the number of clipped speech samples.
9. A method of adjusting a level of a speech signal (400) in a speech communication unit according to Claim 8 - 18 further characterized in that the step of adjusting a gain is performed according to a predetermined relationship or a look-up table function based on the number of clipped speech samples.
10. A speech communication unit (200) substantially as hereinbefore described with reference to, and/or as illustrated by, FIG. 2 of the accompanying drawings.
11. A method of adjusting a level of a speech signal (400) in a speech communication unit substantially as hereinbefore described with reference to, and/or as illustrated by, FIG. 4 of the accompanying drawings.