CN1255255A - Echo reducing phone with state machine controlled switches - Google Patents

Abstract

The purpose of the present invention is to reduce the echo introduced by crosstalk. The above-mentioned problem of how to reduce crosstalk and introduce echo is solved by introducing a switch controlled by a state machine to the microphone and the speaker. The state machine takes the signal energy of the signal from the microphone, the VAD flag of the signal from the microphone, and the The signal energy and the VAD signature of the signal reaching the loudspeaker are taken as input.

Description

Echo reduction phone with state machine controlled switch

This invention relates generally to telecommunications, and more particularly to speech processing for voice communications over the Internet.

A typical Internet phone utilizes a PC with a soundboard, microphone and two speakers. The microphone and speaker are usually placed close to each other on a table. This configuration results in a considerable amount of crosstalk that sounds like an echo at the receiver. This echo must be suppressed in order for Internet telephony to be usable.

In GSM it is known to use VAD (Voice Activity Detection) to detect whether a mobile phone user is speaking or not. Utilizing this information reduces bandwidth when transmitting voice. In discontinuous speech coding according to the VOX principle (Voice Operated Transmission), the VAD unit is responsible for detecting whether the received sound sequence represents human speech. The VAD unit can take two different states, where the first state indicates that the sound sequence is human voice and the other state indicates that the sound sequence is not human voice.

If the VAD unit detects that a given sound sequence represents human speech, the VAD unit will issue a first status signal to the speech encoding unit, which encodes the sound sequence in a speech frame. Conversely, if the given sound sequence represents something other than human speech, the VAD unit will issue a second status signal to the SID (Silence Descriptor) unit. The SID unit sends a SID frame every N frames. Nothing is sent on the remaining N-1 possible occasions to send a frame. The SID frame includes information on the sender about the estimated background noise and the estimated noise spectrum. Using this process saves battery power and radio bandwidth.

When the SID unit changes from generating the first state signal to generating the second state signal, that is the time interval from the detection of speech to the detection of non-speech, usually imposing a so-called release delay, during which time the speech encoding unit continues to send speech frame, it seems that the received sound sequence is still human speech. If the VAD unit still detects non-speech after the release delay time, a SID frame is generated. The reason for this process is that short pauses between words in human language should not be interpreted as non-speech, and the speech frame generator should still be active.

The invention discloses a method and a device for reducing the echo introduced by crosstalk.

It is therefore an object of the present invention to reduce echoes introduced by crosstalk.

The above-mentioned problem of how to reduce the echo introduced by crosstalk is solved by introducing a switch to the microphone and speaker controlled by a state machine, which uses the signal energy of the signal from the microphone, the VAD flag of the signal from the microphone, the signal The signal energy and the VAD flag of the signal to the loudspeaker are taken as input.

One of the advantages of the present invention is that crosstalk-induced echoes are significantly reduced without requiring more computing power.

Other advantages will be apparent to those skilled in the art from the detailed description given below.

Further scope of applicability of the present invention will be apparent from the detailed description given below. It should be understood, however, that the preferred embodiment of the invention is illustrative only, since various changes and modifications within the scope of the invention will become apparent to those skilled in the art from this detailed description.

Figure 1 shows a block diagram of one embodiment of the present invention.

Figure 2 shows a finite state diagram.

In FIG. 1 a microphone 101 is connected to a GSM encoder 102 . Before the signal reaches the GSM encoder 102, it is digitized and sampled according to known techniques not shown in FIG. The encoded signal is first transmitted from the GSM encoder 102 to a receiver (not shown in the figure) via a switch 103 which can activate or deactivate the transmission. _ACFE (Auto Correction Coefficient) is passed from the GSM encoder 102 to the VAD unit 104 . The long-term predictor lag value _NE is also passed to the VAD unit 104 from the GSM frame. A value _PE representing the energy of the signal is passed from the VAD unit 104 to the finite state machine 105 . The VAD unit 104 also calculates a flag _FE indicating whether the VAD unit 104 has detected human speech. The flag _FE is passed to the finite state machine 105 . Flag _FE is true if human voice is detected.

Also shown in FIG. 1 is a sampled encoded speech signal received from a sender (not shown) and passed to the GSM decoder 106 . From the GSM decoder 106 the decoded sampled voice signal is first delivered to the speaker 107 through a switch 108 which enables or disables the voice signal to the speaker. According to known techniques not shown in Fig. 1, a D/A conversion is required for the loudspeaker to function properly. The long-term predictor lag value N ₀ is derived from the received encoded speech signal and passed to the VAD unit 109 .

Since the decoding of GSM frames usually does not involve the use of VAD units, GSM decoders lack the necessary parameters for calculating the ACF. To be able to calculate the ACF, the autocorrelation unit 110 receives the data from the GSM decoder 106 and calculates the ACF _D which is passed to the VAD unit 109 . The autocorrelation unit 110 is part of the GSM encoder described in the standard. An indication value _PD of the energy in the speech signal arriving at the loudspeaker is passed from the VAD unit 109 to the finite state machine 105 . A flag _FD is also passed from the VAD unit 109 to the finite state machine, indicating whether the VAD unit detected human speech or not.

The finite state machine 106 includes the function of setting the switches 103 and 109 according to the values input to the finite state machine.

FIG. 2 shows the states and possible transitions of the finite state machine in FIG. 1 .

Transitions between states are performed as described below. Use the following definitions:

· F _E : VAD flag when encoding

F _D : VAD flag when decoding

P _E : signal energy during encoding

P _D : signal energy when decoding

·Release delay: the time from when the switch direction is determined to when the switch is made. This time must be long enough to compensate for room echoes.

201. _FE = 1AND F _D = 0 OR _FE = 1 and P _E > P _D , release delay = 0

202. F _E = 0, release delay = 600ms

203. F _D ＝1 AND _FE ＝0 OR F _D ＝1 and P _D >P _E , release delay=0

204. F _D = 0, release delay = 600ms

205.F _D ＝1 AND P _D ＞P _E , release delay＝600ms

_205.FE = 1 AND P _E > P _D , release delay = 600ms

In state TRANSMITTING 207, the switch controlling the transmission of the voice signal from the microphone is activated and the switch controlling the transmission of the voice signal to the speaker is disabled. In state RECEIVING 208, the switch controlling the transmission of voice signals from the microphone is disabled and the switch controlling transmission to the speaker is enabled. In the IDLE state 209, both switches are off.

The invention thus being described, it will be obvious that the invention may be varied in various ways. While such changes are not to be regarded as a departure from the spirit and scope of the invention, all such modifications as would be obvious to one skilled in the art are intended to be embraced within the scope of the following claims.

Claims (9)

1. method that is used for when the phone application transporting speech reducing echo, described phone application comprises loud speaker and microphone, it is characterized in that, finite state machine influences opening or closing of described loud speaker and microphone according to the signal characteristic that reaches described loud speaker from the signal characteristic of described microphone.

2. according to the method for claim 1, wherein said phone application comprises at least one VAD unit, a GSM encoder and a GSM decoder, it is characterized in that, to pass to described finite state machine from the VAD sign of the signal of microphone will represent to pass to described finite state machine from first value of the representation signal energy in the signal of microphone, the 2nd VAD sign that arrives the signal of loud speaker is passed to described finite state machine, second value of the energy in the signal of expression arrival loud speaker is passed to described finite state machine, according to described VAD sign, described the 2nd VAD sign, described first value and described second value, described finite state machine influence control is from first switch of the transmission of the described signal of described microphone, and described finite state machine is transferred to described secondary signal the second switch of described loud speaker.

3. according to the method for claim 2, it is characterized in that, to pass to described GSM encoder from the first sampling voice signal of described microphone, the first long-term predictor lagged value is passed to a VAD unit, to pass to a described VAD unit from first auto-correlation coefficient of a described GSM encoder, to pass to described finite state machine from first Boolean denotation of a described VAD unit, expression is passed to described finite state machine from first value of the energy of the signal of described microphone from a described VAD unit, receive the second sampling coded speech signal, described second voice signal is passed to the GSM decoder, to pass to the 2nd VAD unit from the second long-term predictor lagged value of described second voice signal, calculate second auto-correlation coefficient and pass to described the 2nd VAD unit, second value of the energy in described second voice signal of expression is passed to described finite state machine from described the 2nd VAD unit, second Boolean denotation is passed to described finite state machine from described VAD unit, and described finite state machine is according to described first Boolean denotation, described second Boolean denotation, described first value and described second value, the control influence is from first switch of the transmission of the described first sampling coded speech signal of described microphone, and the described second decoding voice signal of influence is to the second switch of the transmission of loud speaker.

4. according to the method for claim 2, it is characterized in that, if described finite state machine is got first state, control is arranged to allow this transmission from described first switch of the transmission of described microphone, and control transmission is arranged to not allow this transmission to the described second switch of loud speaker, if described finite state machine is got second state, control is arranged to not allow this transmission from described first switch of the transmission of described microphone, and will controls the described second switch of the transmission of loud speaker is arranged to allow this transmission.

5. according to the method for claim 4, it is characterized in that,, then described first and second switch all is arranged to identical state if described finite state machine is got the third state.

6. according to the method for claim 5, it is characterized in that, if if described first be masked as true with described second be masked as puppet or described first be masked as pseudo-and described first value greater than described second value, then described finite state machine switches to described first state from the described third state; If described first is masked as puppet and has pass by hang-over delay, then described finite state machine switches to the described third state from described first state; If if described second be masked as true and described first and be masked as puppet or described second and be masked as true and described second value greater than described first value, then described finite state machine switches to described second state from the described third state; If described second is masked as pseudo-and has pass by described hang-over delay, then described finite state machine switches to the described third state from described second state; If described second be masked as true and described second value greater than described first value with pass by described hang-over delay, then described finite state machine switches to described second state from described first state; If described first is masked as true and described first value greater than described second value and pass by described hang-over delay, then described finite state machine switches to described first state from described second state.

7. according to the method for claim 6, it is characterized in that described hang-over delay is 600ms.

8. device that is used for when the phone application transporting speech reducing echo, described phone application comprises loud speaker and microphone, it is characterized in that described phone application comprises the finite state machine that opens or closes that is configured to according to influence described loud speaker and microphone from the signal characteristic of described microphone and the signal characteristic that arrives described loud speaker.

9. one kind is configured to phone application transport and the personal computer that receives speech, described phone application comprises the echo that is used to reduce described speech, described phone application comprises loud speaker and microphone, it is characterized in that described phone application comprises that basis reaches the finite state machine that opens or closes that influences described loud speaker and microphone from the signal characteristic of described loud speaker from the signal characteristic of described microphone.

Images