Hand-free telephone set

p0001A hands-free telephone unit allows the user to converse with a remote party without holding the handset in hand.

p0002A handsfree set includes a microphone and a speaker embedded in its case, more sensitive than those of the handset. Thus, the user can use his post-hook by being distant.

p0003This increase in the scope for the job emission path which is the microphone and the receive path leading to the speaker, however, induced technical problems because the voice signal from the caller is returned to high level by the speaker and undergoes reflection on the walls of the room where the post as well as the position of the box, before being partially in turn captured by the microphone. The captured spurious signal is sent online, superimposed on the voice signal of the local user, and provides for the corresponding echo.

p0004Moreover, as the phone line transmits signals in both directions, the station includes a duplexer circuit, line interface, connecting the line to the transmission channel and the position of the reception path. The duplexer allows injection line signals to transmit and receive the signals from the corresponding, but not receive the signals injected locally to avoid a clash sound or howling. In fact, on the receive path, the duplexer can not avoid taking a composite signal representing all of the received and transmitted signals, but after sampling the receive path, it subtracts the signal from the microphone so n ' apply to the speaker that the signal received from the correspondent. This subtraction or balancing, canceling the return of the transmitted signals, however, is never perfect because the balance referred depends on the line impedance, which can vary from one line to another and over time. Therefore, in practice, part of the signal transmitted from the microphone is reflected, at the duplexer to the speaker.

p0005In short, the transmission and reception of the station assault is a sound loop, comprising, on one side, the acoustic reflection of the local and on the other hand, the electrical reflection of the duplexer. Or, in the case of hands-free position, sensitivity microphone and loudspeaker such as browsing the loop signal is amplified to saturation circuits and producing a feedback.

p0006To avoid feedback and echo effect mentioned above, it is known to run the post-duplex, that is to say only in transmission or reception, cutting the Larsen hanging loop and echo. Continuous measurement of the microphone signal and the received line to determine which of the two interlocutors that produces the strongest vocal energy and activate accordingly in the position, the transmit path or the receive path.

p0007Such a solution, however, has drawbacks as a talkative interlocutor and loudly long monologue can not be switched off by the other. In addition, if it is two hands-free stations are connected, their operation in alternating is not synchronized, so that they may be in phase, that is to say both in transmission, no reception, or reception, with zero emissions. This leads to a chopped conversation.

p0008It can be expected to make anti-echo filtering, acoustic and electric, using adaptive filters, that is to say, gradually changing their filter settings by determining a model of local or online to converge the energy echo to a minimum. Acoustic feedback may also be removed by a filter performing a frequency shift, in case of detection of acoustic coupling on one frequency, and decreasing the loop gain thereof.

p0009However, the filter convergence time are relatively high and fail to adapt to rapid environmental changes, such as moving a person or opening a door modifying local echoes.

p0010The present invention aims to overcome the disadvantages mentioned above, to enable full-duplex operation.

p0011To this end, the invention relates to a hands-free telephone set arranged to be connected to a telephone line carrying sound signals and comprising a transmission channel and a reception channel, each channel comprising means for attenuating and adaptive filtering means of echoes of the other channel, station characterized in that it comprises means for measuring an adaptation defect in the filtering means, the measuring means being arranged to activate the means of attenuation of at least one channel when the defect exceeds a predetermined threshold.

p0012Thus, from the appearance of a defect exceeding a threshold, activation of mitigation measures passed the full-duplex operation, without mitigation, an operating-duplex or full-duplex degraded, with mitigation. This stifles the effects of mismatch and thus allows filtering means to rehabilitate in good conditions, so then eliminate the defect and allow the measurement means to restore the full-duplex operation when the new environment been fully understood.

p0013The invention will be better understood using the following description of a preferred embodiment of the handsfree set of the invention, with reference to the accompanying drawing, in which:<ul><li>Figure 1 is a functional block of the position of the invention, and</li><li>2 is a flowchart illustrating the sequence of sound adjustment of the position.</li></ul>

p0014The hands-free station shown includes a transmission channel 1 and channel 2 receiving, connected to a telephone line 10, and a central unit 3 commander tracks 1 and 2. In general, the phone line carrying sound signals is any link for transmitting a conversation, such as a radio link or a data transmission network.

p0015Transmitting path 1 comprises, in the direction of transmission of sound signals, a microphone 11 for hands-free, an analog / digital converter (ADC) 12 connected to first inputs of an adder-subtracter 13 itself associated an acoustic echo cancellation filter 14 mounted in loop between its output and its second inputs, an anti-feedback filter 15, a controlled attenuator 16 and a digital / analog converter (DAC) 17 whose output is connected to line 10 through a hybrid circuit 20 passage 2 son (line 10) / 4 son (2 emission son, 2 son receipt).

p0016The receive path 2 comprises, from the circuit 20, an ADC 22 connected to first inputs of an adder-subtracter 23 itself conjunction with an anti-echo filter 24 mounted in line loop between its output and its second inputs , a controlled attenuator 26 and a DAC 27 controls a speaker 28 of hands-free.

p0017The filter 14 also receives the signal from the attenuator 26, while filter 24 receives the attenuator 16.

p0018The CPU 30 includes a measurement circuit 31 of energies of the injected signals on channels 1 and 2, connected by the output of the adders-subtracters 13 and 23. A circuit 32 for measuring mismatches, comprising a microprocessor, receives the measurement results above, as well as fault signals or mismatch, from the filters 14 and 24.

p0019The control circuit 32, by not all shown connecting the circuits 13 to 16 and 23 to 26 as explained later. It also provides a conventional manner, synchronization of ADC and DAC 12, 22 and 17, 27 in relation to other above circuits 13-16 and 23-24, 26, all of which are digital.

p0020Although an analog processing of signals of the two channels is not excluded, thus without need of ADC and DAC, is here a single signal processing processor which digitally performs the functions of the circuits 13-16 and 23-24 26, the ADC and DAC 12, 22 and 17, 27 can also be integrated into the processor. It could also have been provided a digital line.

p0021The operation of the station will now be explained with reference to the flowchart of Figure 2.

p0022Under the control of the measurement circuit 32, channels 1 and 2 pass through various states based on information from the circuits monitored by the circuit 32.

p0023When establishing a telephone call, at step 41, the circuit 32 initializes, that is to say set to zero, the filter coefficients used in filter 14 to model the local acoustically in which is the station shown. The circuit 32 also controls the insertion of attenuation, here on the two channels 1 and 2 by means of the attenuators 16 and 26. The attenuation in at least one location of the phonic loop consisting of channels 1 and 2, the circuit 20 and the walls of the room and avoids hooking Larsen. In addition, the attenuation of the transmission channel 1 will prevent a return of excessive acoustic echo to the line 10.

p0024In a step 42, the circuit 32 controls the filter coefficients 14 to freeze, that is to say freeze the acoustic model of the local after a certain period of adjustment, or, upon detection of absence of mismatch as explained below. It also controls the initialization or reset of the line filter coefficients 24 and the activation of the adaptation thereof to model the line 10 and the hybrid circuit 20.

p0025We then go to step 43 in which the circuit 32 monitors a mismatch line 24 filter.

p0026It will be recalled that an anti-echo filter temporarily stores the signal generating the echo and then subtract it to the other channel signal affected by this echo. The adaptation is to delay and weighting, appropriately, each frequency component of the signal generating the echo to subtract the echo signal. If imperfect adjustment of determining filter coefficients adaptation, the filter detects that there remains a filtering residue, or remaining echo having a correlation with the signal generating echo and it develops its coefficients to try to cancel this echo by converging its model to the real model of the room. The filter residue represents a mismatch, which is transmitted to the circuit 32.

p0027For details on the filters 14, 15, 24, reference is made to US-A-5099472, WO 89/05068, JP 5-14476 and JP 5-48707, incorporated herein by reference.

p0028The circuit 32 monitors here, at step 43, the amplitude of the above fault line 24 from the filter, if it exceeds a predetermined threshold, back to step 42 and thus reset the line filter 24 to avoid possible divergence, delaying or even preventing an adaptation, or convergence, later. Similarly, a slow convergence is detected, that is to say, we also returns to step 42 if a predetermined threshold time of non-adaptation is exceeded. A mismatch thus corresponds to the exceeding of a threshold amplitude of the mismatch signal or the exceeding of a time threshold during which the mismatch signal exceeds a predetermined value.

p0029At all stages, the circuit 31 ensures the detection of an acoustic feedback and the circuit 32 then controls the filter 15 which then shifts slightly the signal frequency spectrum therethrough and thereby immediately reduces the loop gain for the frequency of Larsen coupling.

p0030Without returning to the step 42, the circuit 31 detects, at step 44, if the energy is present on a single channel, 1 or 2, and informs the circuit 32, which disables the attenuator 16 or 26 of the other channel, 2 or 1. in case of presence of power on the two channels 1 and 2, the circuit 32 disables the attenuator 16 or 26 of the channels 1 and 2 which carries the signal lower energy.

p0031At step 45, the adjustment is continued the line pattern in the filter 24 if the circuit 31 detects a transmission channel energy signal 1. is continued further comparing the energies of the two channels 1 and 2 for the circuit 32 switches the attenuation on the tracks which the lower energy signal. It works well in pseudo-duplex, maintaining a low level transmission on the way "off."

p0032In a following step 46, the adaptation of the line model in the filter 24 has continued to achieve a satisfactory level of echo cancellation line. If the receive channel 2 receives a voice signal from a remote party and the local echo is dominant in the transmission channel 1, the signal of the extension user, the circuit 32 controls the adaptation of the local model in the filter 14.

p0033At a following step 47, counterpart of step 43, it is detected whether the filter convergence time 14 or the magnitude of the mismatch signal exceeds a predetermined threshold and then returns to step 46 after resetting coefficients of filter 14.

p0034In the absence of mismatch on each of the transmission and reception paths 1 and 2, it goes to a step 48 in which the circuit 32 disables the attenuator circuits 16, 26 that was active and thus establishes a state of true full-duplex without mitigation.

p0035It is understood that by "without mitigation" means a substantially consistent transmission amplitude template provided for the transmission of sound signals through the telephone network, to provide or restore a sound signal of sufficient amplitude but without risking saturation circuits of the telephone network or transceiver channels of the post. To adjust the paths above as such a template, it may be necessary to continuously provide them attenuation, which does not concern the variable attenuation of the present invention.

p0036In step 48, it continues to adapt models of filters 14 and 24. In case of occurrence of a mismatch of the model (14) of the local exceeding a low threshold, here below the threshold (high ) in step 47, the circuit 32 applies (step 49) limited attenuation (16) on the transmission channel 1. in other words, and in general, 32 the circuit activates the attenuator 16 when the defect exceeds the low threshold and control an increase in mitigation when the upper threshold is exceeded. It returns to step 48 when the low threshold of step 49 is crossed descent.

p0037In case of exceeding the mismatch predetermined high threshold filter 14, it returns to step 46, that is to say the 32 circuit activates one of the attenuators 16, 26, here the attenuator 16 of the transmission channel 1 to return the intermediate operation between full-duplex and half-duplex. We go here by step 47 to reset the filter coefficients 14.

p0038It could be provided to enable the two attenuators 16, 26, as in step 41.

p0039In other words, in steps 41 to 48, the circuit 32 disables here, in case of defect exceeding the threshold, the filters 14, 24 of the two channels 1 and 2 in succession one after the other until no from defects exceeding the threshold. This sets more accurately each filter 14, 24, avoiding a change in the other filter 24, 14. In particular, the filter 14 of the transmit path 1 is turned off (step 42) before the filter 24 of the receive path 2.