Summary of the invention
Purpose of the present invention is exactly above two problems that are primarily aimed at prior art, proposes a kind of echo processing method of TV conference system.
For this reason, the echo processing method of TV conference system of the present invention comprises the steps: A, carries out the state judgement according to audio sampling data, and is transformed into corresponding state; B, according to the audio-frequency processing method of predefined corresponding state, audio frequency is handled.
Preferably, the present invention also comprises following feature:
State described in the described steps A comprises mute state, and promptly the absolute value of this meeting-place all audio sampling datas in the continuous time is all less than the situation of first threshold values; Among the step B under this state the disposal route to audio frequency be: aftertreatment is carried out in the output to echo cancellation module, replaces the output of echo cancellation module with quiet data.
State described in the described steps A comprises attenuation state, and promptly the absolute value of this meeting-place all audio sampling datas in the continuous time is all less than the situation of second threshold values; Among the step B under this state the disposal route to audio frequency be: pre-treatment is carried out in the input to echo cancellation module, and audio sampling data is decayed half, and then sends into echo cancellation module and handle.
The value of M1 is not make a speech when the both sides meeting-place, under the very quiet situation in this meeting-place, and the value of absolute value maximum in the A/D output valve.
The value of M2 M2 is when meeting-place, opposite end speech, and do not make a speech under the only reflective situation in this meeting-place, the value of absolute value maximum in the A/D output valve.
Described state also comprises floor status, promptly in the sampling time of regulation the absolute value of not every audio sampling data all less than first threshold values (M1) or all less than the situation of second threshold values (M2); Transformation rule between each state is: when the meeting-place is in mute state, if detect certain audio sampling data more than or equal to M2, then enter floor status; When the meeting-place is in mute state,, then enter attenuation state if detect certain audio sampling data more than or equal to M1 and less than M2; When the meeting-place is in attenuation state,, then enter floor status if detect certain audio sampling data more than or equal to M2; When the meeting-place is in attenuation state,, then enter mute state if detect 1 audio sampling data of N continuous less than M1; When the meeting-place is in floor status,, then enter mute state if detect 1 audio sampling data of N continuous less than M1; When the meeting-place is in floor status,, then enter attenuation state if detect 2 audio sampling datas of N continuous less than M2; When N1=N2, the meeting-place is in floor status, if detect 2 audio sampling datas of N continuous less than M2, satisfies 1 audio sampling data of N continuous so simultaneously less than M1, enters mute state this moment; Wherein:
N1=t1 * audio sample rate, N2=t2 * audio sample rate.
This pre-treatment scheme that can improve echo neutralization effect proposed by the invention, by judging the current state in meeting-place, and to the pre-treatment of audio sampling data and the aftertreatment that echo cancellation module is exported, reach better echo neutralization effect, the situation of speech simultaneously not in the both sides meeting-place can be improved the effect of echo cancelltion effectively.
Embodiment
This programme according to the current state of judging the meeting-place, reaches desirable echo neutralization effect by the aftertreatment of audio sampling data being carried out pre-treatment and echo cancellation module is exported by the method for software.As shown in Figure 1, its A/D represents the analog-to-digital conversion module of audio sampling data.
Present embodiment is divided into following three kinds with the state in meeting-place, judges current conference site state and handles accordingly by software.
(1) state 1: mute state
When the absolute value of this meeting-place all audio sampling datas in continuous time t1 all under the situation less than M1, we think that this meeting-place is in mute state.This state normally the microphone in this meeting-place close or microphone open but do not have the speech and the very little situation of echo.Under this kind state, we carry out aftertreatment to the output of echo cancellation module, with the output of quiet data (referring to allow the D/A of opposite end export the data of " 0 ") replacement echo cancellation module, to reach extraordinary echo neutralization effect in the meeting-place, opposite end.
(2) state 2: attenuation state
When the absolute value of this meeting-place all audio sampling datas in continuous time t2 all under the situation less than M2, we think that this meeting-place is in attenuation state.This state normally microphone in this meeting-place is opened but is not had speech and the bigger situation of echo.Under this kind state, we carry out pre-treatment to the input of echo cancellation module, and audio sampling data is decayed half, and then send into echo cancellation module and handle, thereby obtain desirable echo neutralization effect in the meeting-place, opposite end.
(3) state 3: floor status
When the condition that requires in state 1 and the state 2 can not be satisfied in this meeting-place (absolute value of not every audio sampling data is all less than M1 or all less than M2 in the sampling time of regulation), we think that this meeting-place is in floor status.At this moment, we do not carry out special processing.
Below present embodiment is explained in more detail:
One, choosing of threshold values:
In the present embodiment, threshold values M1 and M2 choose the judgement that is related to current conference site state, thereby influence follow-up processing, and be therefore extremely important.
Audio sampling data has just to be had negatively, is greater than or less than certain threshold values in the present embodiment and is meant that all its absolute value is greater than or less than certain threshold values (other places explanation).
M1 in the present embodiment and M2 must satisfy following relation: 0<M1<M2.
The value of M1 is not make a speech when the both sides meeting-place, under the very quiet situation in this meeting-place, and the value of absolute value maximum in the A/D output valve.
The value of M2 is when meeting-place, opposite end speech, and do not make a speech under the only reflective situation in this meeting-place, the value of absolute value maximum in the A/D output valve.
Generally, in different meeting-place, the value of M1 and M2 is inequality.But it is general all in certain scope.Therefore we that these two threshold values are designed to program is variable, promptly, be set to different values by application program according to different meeting-place environment.
Two, the change of state: as shown in Figure 2---
1. when the meeting-place is in state 1,, then get the hang of 3 if detect certain audio sampling data more than or equal to M2.
2. when the meeting-place is in state 1,, then get the hang of 2 if detect certain audio sampling data more than or equal to M1 and less than M2.
3. when the meeting-place is in state 2,, then get the hang of 3 if detect certain audio sampling data more than or equal to M2.
4. when the meeting-place is in state 2,, then get the hang of 1 if detect 1 audio sampling data of N continuous less than M1.
5. when the meeting-place is in state 3,, then get the hang of 1 if detect 1 audio sampling data of N continuous less than M1.
6. when the meeting-place is in state 3,, then get the hang of 2 if detect 2 audio sampling datas of N continuous less than M2.
7. when N1=N2, the meeting-place is in state 3, if detect 2 audio sampling datas of N continuous less than M2, because N2=N1, and M1<M2, so satisfy 1 audio sampling data of N continuous simultaneously less than M1, get the hang of 1 this moment.
Wherein:
N1=t1 * audio sample rate
N2=t2 * audio sample rate
For example we think that when all audio sampling data is all less than M1 in continuous 1 second be mute state, if audio sample rate is 8K, then N1 is 8000.The value of N2 in like manner.
Three, treatment scheme: shown in Fig. 3-7
Fig. 3 is the present embodiment schematic flow sheet.Wherein, T is an audio sampling data, and original state is set is state 3.After receiving audio sampling data T, relatively judge with threshold values more as calculated, and carry out state exchange according to judged result.Its transformation rule as mentioned above.Its flow path switch as shown in Figure 4.
After being transformed into arbitrary state, just handle by the tupe of this state.Fig. 5 is the treatment scheme synoptic diagram of present embodiment to state 1.Wherein, Cnt1, Cnt2 are integer variable (int), be used for counting, and to make the initial value of Cnt1 and Cnt2 are 0.Fig. 6 then is the treatment scheme synoptic diagram of present embodiment to state 2.Fig. 7 then is the treatment scheme synoptic diagram of present embodiment to state 3.
Processing scheme of the present invention can reach better echo neutralization effect by to the pre-treatment of audio sampling data and the aftertreatment that echo cancellation module is exported.