AU735505B2

AU735505B2 - Echo reducing phone with state machine controlled switches

Info

Publication number: AU735505B2
Application number: AU64264/98A
Authority: AU
Inventors: Johan Gnosspelius
Original assignee: Telefonaktiebolaget LM Ericsson AB
Current assignee: Telefonaktiebolaget LM Ericsson AB
Priority date: 1997-03-11
Filing date: 1998-02-24
Publication date: 2001-07-12
Anticipated expiration: 2018-02-24
Also published as: WO1998040974A1; JP2001514823A; TW407435B; SE511650C2; CA2283590A1; CN1255255A; EP0974205A1; SE9700873L; BR9808240A; AU6426498A; SE9700873D0

Description

WO 99140974 PCT/SE98/00332 1 ECHO REDUCING PHONE WITH STATE MACHINE CONTROLLED

SWITCHES

TECHNICAL FIELD OF INVENTION The present invention relates to telecommunication in general and to speech processing for voice communication over the Internet in particular.

DESCRIPTION OF RELATED ART A typical Internet phone uses a PC with a sound board, a microphone and two loudspeakers. The microphone and the loudspeakers are often placed next to each other on the desk.

Such a configuration causes a considerable amount of cross-talk heard as an echo at the receiver end. This echo must be suppressed to make the Internet phone usable.

In GSM it is known to use VAD (Voice Activity Detection) to detect when a user of a mobile phone is talking or not talking.

This information is used to be able to decrease to bandwith when transmitting the voice. In discontinuous speech coding according to the VOX-principle (Voice Operated Transmisson) a VAD unit is responsible for detecting wheter or not a received sound sequence represents human speech or not. The VAD unit can take two different states where a first state indicates that the sound sequence was human voice and the other state indicates that the sound sequence was not human voice.

If the VAD unit detects that a given sound sequence represents human voice the VAD unit will issue a first state signal to a speech coding unit which will encode the sound sequence in a speech frame. If on the other hand a given sound sequence represents something but not human speech the VAD unit will issue a second state signal to a SID (Silence Descriptor) unit.

Said SID unit will every N:th frame deliver a SID frame. During the remaining N-l possible occasions to send frames nothing will SUBSTITUTE SHEET (RULE 26) I, i. l: -r 2 be sent. A SID frame comprises information about estimated background noise and estimated noise spectra on the sending side. With this procedure battery power and radio-bandwidth can be saved.

When the SID unit changes from generating the first state final to generating the second state signal, that is from detecting speech to detecting non-speech a time interval, a so called hangover is normally applied during which the speech encoding unit will continue to deliver speech frames as if the received sound sequence had been human speech. If, after the hang-over time the VAD unit still detects non-speech a SID frame is generated. The reason for this procedure is that short pauses between words in the human speech not shall be interpreted as non-speech, but that the speech frame generator still shall be active.

SUMMARY OF THE INVENTION The present invention discloses a method and an apparatus for reduction 15 of echoes introduced by cross-talk.

The purpose of the present invention is thus to reduce the echo introduced by cross-talk.

The present invention accordingly provides a method for reducing echo when transmitting voice in a telephone application, said telephone application including a speaker and a microphone, characterised in that a finite state machine affects said speaker and microphone to be on or off in dependence of characteristics of the signal from said microphone and characteristics of the signal to said speaker. According to another aspect, the invention provides apparatus for reducing echo when transmitting voice in an telephone application, said telephone application including a speaker and a microphone, characterised in that said telephone application includes a finite state machine arranged to affect said speaker and microphone to be on or off in dependence of characteristics of the signal from said microphone and characteristics of the signal to said speaker.

Accordingly to yet another aspect, the invention provides a personal computer arranged for transmitting and receiving voice with an telephone application AFniA/ipcluding an apparatus for reducing echo of said voice, said telephone application ~li- I; r. rL 1. ;;il.r lc ,llr* including a speaker and a microphone, characterised in that said telephone application includes a finite state machine arranged to affect said speaker and microphone to be on or off in dependence of characteristics of the signal from said microphone and characteristics of the signal to said speaker.

One of the advantages with the present invention is that the echo introduced with cross-talk is significantly reduced without requiring too much computational power.

Other advantages will be obvious to a man skilled in the art in the light of the detailed description given below.

Further scope of applicability of the present invention will become apparent from the detailed description given herein after. However, it should be understood that the preferred embodiments of the invention, are given by way of illustration only, since various changes and modifications with the scope of the invention will become apparent to those skilled in the art from this detailed 15 description.

i BRIEF DESCRIPTION OF DRAWINGS Figure 1 shows a block diagram of one embodiment of the invention.

Figure 2 shows a Finite State Diagram.

DETAILED DESCRIPTION OF PREFERRED EMBODIMENTS S 20 In Figure 1 a microphone 101 is connected to a GSM encoder 102. Before "the signal arrives to the GSM encoder 102 it has been digitalised and sampled according to known technology but which is not shown in figure 1. From the GSM o encoder 102 the encoder signal is transmitted to a receiver not shown in the lO.t figure first passing a switch 103 which can enable or disable the transmission.

From the GSM encoder 102 is a ACFE (Autocorrection CoeFficient) passed to a VAD unit 104. To the VAD unit 105 is also a long term predictor lag value NE is a value, PE representing the energy of the signal passed to a finite state machine 105. The UA APN

CI

TO0' WO 98/40974 PCT/SE98/00332 4 VAD unit 104 also computes a flag FE indicating if the VAD unit 104 has detected human speech. The flag FE is passed to the finite state machine 105. The flag FE will be true if human voice has been detected.

Further in figure 1 is a sampled, encoded voice signal received from a sender (not shown) and passed to a GSM decoder 106. From the GSM decoder 106 the decoded, sampled voice signal is passed to a speaker 107 first passing a switch 108 which can enable or disable to voice signal from reaching the speaker. For the speaker to be able to function properly an D/A-convertion is needed according to known technology but not shown in figure 1.

From the received sampled, encoded voice signal a long term predictor lag value ND is deducted and passed to a VAD unit 109.

Since the decoding of GSM frames normaly do not involve using a VAD unit the GSM decoder lacks necessary parameters for calculating the ACF. To be able to calculate the ACF an autocorrelation unit 110 receives data from the GSM decoder 106 and calculates the ACFD which is passed to the VAD unit 109. The autocorrelation unit 110 is a part of the GSM encoder as described in the standards. A value PD indication the energi in the voice signal to the speaker is passed from the VAD unit 109 to the finite state machine 105. From the VAD unit 109 is also a flag FD passed to said finite state machine indicating if the VAD unit has detected human voice.

The finite state machine 106 comprises functionality for setting the switches 103 and 109 in dependance of the values inputted to the finite state machine.

In figure 2 the states and the possible transitions is shown for the finite state machine in figure 1.

SUBSTITUTE SHEET (RULE 26) i. 1..1 The transitions between states are done according to the description below. The following definitions are used: FE: VAD flag when encoding FD: VAD flag when decoding PE: Signal energy when encoding PD: Signal energy when decoding.

Hangover: The time from the decision to switch direction until the switch is made. This time must be long enough to compensate for the room echo.

201. FE= 1 AND FD= 0 OR FE= 1 and PE> PD, hangover 0 202. FE= 0' hangover 600ms 203. FD= 1 AND FE= 0 OR FD= 1 and PD> PD, hangover 0 204. FD= 0, hangover 600 ms 205. FD= 1 AND PD PE, hangover 600 ms 206. FE= 1 AND PE PD, hangover 600 ms 15 In the state TRANSMITTING 207 is the switch controlling the transmission of the voice signal from the microphone enabled and the switch controlling the transmission of the voice signal to the speaker disabled. In the state RECEIVING 208 is the switch controlling the transmission of the voice signal from the microphone disabled and the switch controlling the transmission to the speaker 20 enabled. In the IDLE state 209 both switches are disabled.

The invention being thus described, it will be obvious that the same may be varied in many ways. Such variations are not to be regarded as a departure from the spirit and scope of the invention, and all such modifications as would be obvious to a man skilled in the art are intended to be included within the scope of the following claims.

i L 1 1. .l.i;

Claims

1. A method for reducing echo when transmitting voice in a telephone application, said telephone application including a speaker and a microphone, characterised in that a finite state machine affects said speaker and microphone to be on or off in dependence of characteristics of the signal from said microphone and characteristics of the signal to said speaker.

2. A method according to claim 1 where said telephone application includes at least one VAD unit, one GSM encoder and one GSM decoder, characterised in that a first VAD flag of the signal from the microphone is passed to said finite state machine, that a first value representing the signal energy in the signal from the microphone is passed to said finite state 15 machine, that a second VAD flag of the signal to the speaker is passed to said S finite state machine, that a second value representing the energy in the signal to the speaker is passed to said finite state machine, that said finite state machine affects a first switch controlling the transmission of said signal from said microphone, that said finite state machine affects a.second switch 20 controlling the transmission of said second signal to said speaker in o oo dependence of the values of said first VAD flag, said second VAD flag, said g first value and said second value.

3. A method according to claim 2, characterised in that a first sampled voice signal from said microphone is passed to said GSM encoder, that a first long term predictor lag value is passed to a first VAD unit, that a first autocorrelaiton coefficient is passed from said first GSM encoder to said first VAD unit, that a first boolean flag is passed from said first VAD unit to said finite state machine, that a first value representing the energy of the signal from said microphone is passed from said first VAD unit to said finite state Smachine, that a second sampled, encoded voice signal is received, that said I second voice signal is passed to a GSM decoder, that a second long term predictor lag value from said second voice signal is passed to a second VAD unit, that a second autocorrelation coefficient is calculated and passed to said second VAD unit, that a second value representing the energy in said second voice signal is passed from said second VAD unit to said finite state machine, that a second boolean flag is passed from said VAD unit to said finite state machine and that said finite state machine controls a first switch affecting the transmission of said first sampled, encoded voice signal from said microphone, a second switch affected the transmission of said second, decoded voice signal to a speaker in dependence of the values of said first boolean flag, said second boolean flag, said first value and said second value.

4. A method according to claim 2, characterised in that if said finite state *0oo machine takes a first state and first switch controlling the transmission from 15 said microphone is set to allow such transmission and that said second switch *4, controlling the transmission to the speaker is set not to allow such transmission, that if said finite state machine takes a second state said first switch controlling the transmission from said microphone is set not to allow such transmission and that said second switch controlling the transmission to S 20 the speaker is set to allow such transmission. ,o 5. A method according to claim 4, characterised in that if said finite state machine takes a third state said first and second switch are both set to the Ssame state.

6. A method according to claim 5, characterised in that said finite state machine switch from said third state to said first state if said first flag is true and said second flag is false or if said first flag is true and said first value is larger than said second value, that said finite state machine switches from said first state to said third state if said first flag is false and a hangover time .KSTF& 7 has elapsed, that said finite state machine switches from said third state to i i- l -r L. L .II_-i.ll. 8 said second state if said second flag is true and said first flag is false or if said second flag is true and said second value is larger than said first value, that said finite state machine switches from said second state to said third state if said second flag is false and said hangover time has passed, that said finite state machine switches from said first state to said second state if said second flag is true and said second value is larger than said first value and said hangover time has passed, that said finite state machine switches from said second to said first state if said first flag is true and said first value is larger than said second value and said hangover time has passed.

7. A method according to claim 6, characterised in that said hangover time is 600 ms. C *r a a.. a.. ar. a a p CI a a a I a

8. Apparatus for reducing echo when transmitting voice in an telephone application, said telephone application including a speaker and a microphone, characterised in that said telephone application includes a finite state machine arranged to affect said speaker and microphone to be on or off in dependence of characteristics of the signal from said microphone and characteristics of the signal to said speaker.

9. A personal computer arranged for transmitting and receiving voice with an telephone application including an apparatus for reducing echo of said voice, said telephone application including a speaker and a microphone, characterised in that said telephone application includes a finite state machine arranged to affect said speaker and microphone to be on or off in dependence of characteristics of the signal from said microphone and characteristics of the signal to said speaker. A method according to any one of claims 1 to 7, substantially as described with reference to the drawings. r~L~ r- r~ II_ ii i II I Li I

11. Apparatus according to claim 8, substantially as described with reference to the drawings.

12. A personal computer according to claim 9, substantially as described with reference to the drawings. DATED this 1st Day of May, 2001 TELEFONAKTIEBOLAGET LM ERICSSON (Publ) WATERMARK PATENT TRADE MARK ATTORNEYS Unit 1, "The Village" Riverside Corporate Park

39-117 Delhi Road, North Ryde NSW 2113 o oo *o *•*oo *ooo *o° *2 -I I I 1