WO2004070969A1

WO2004070969A1 - Echo cancellation device including a double talk detector

Info

Publication number: WO2004070969A1
Application number: PCT/IB2003/000667
Authority: WO
Inventors: Dragan Topalovic
Original assignee: Telefonaktiebolaget Lm Ericsson (Publ)
Priority date: 2003-02-03
Filing date: 2003-02-03
Publication date: 2004-08-19
Also published as: EP1590894A1; AU2003206046A1

Abstract

This echo cnacellation device, for cancelling an echo signal (S(n)) output by telecommunication means (10), comprises adaptive filtering means (11), for outputting an estiamte (S'(n)) of the echo signal (S(n)). It further comprises further adaptive filtering means (34), receiving as an input a signal (Y(n)) which comprises the sum of a signal (V(n)) originating from a first end of a connection and of the echo signal (S(n)). The adaptive filtering means (11) and the further adaptive filtering means (34) are controlled by a first error signal (E'(n)), and the first error signal (E'(n)) becomes null if, either the signal (V(n)) originating from the first end of the connection, or the echo signal (S(n)), is suppressed by the further adaptive filtering means (34) and if the estimate (S'(n)) of the echo signal (S(n)) is equal to the echo signal (S(n)).

Description

ECHO CANCELLATION DEVICE INCLUDING A DOUBLE TALK DETECTOR

The present invention relates to an echo cancellation device including a double talk detector. Echo is a problem related to the perceived speech quality in telephony systems with long delays, e.g. telephony over long distances or telephony systems using long processing delays, like digital cellular systems. The echo arises in the four-to-two wire conversion in the PSTN (Public Switched Telephone Network)/subscriber interface. To remove this echo, echo cancellers are usually provided in transit exchanges for long distance traffic, and in mobile services switching centres for cellular applications.

Due to the location of the echo canceller, it is made adaptive; the same echo canceller is used for many different subscribers in the PSTN. This adaptation is necessary not only between different calls, but also during each call, due to the non-fixed nature of the transmission network, e.g. phase slips, three-party calls, etc.

The main part of an echo canceller is an adaptive filter. The filter generates a replica of the echo, which is subtracted from the near signal. Due to imperfect knowledge of the echo generating system, the estimated echo signal always contains errors. Hence, in practice, the echo attenuation obtained by using an adaptive filter is usually at most approximately 30 dB. For long time delays, this attenuation is not enough, and in order to minimise the audible effects of these errors, a residual echo suppressor is used. The purpose of the echo suppressor is to further suppress the residual signal whenever this signal is dominated by the errors in the echo estimate. Blocking the output of the echo canceller for certain levels of the output signal does this.

Many solutions have been proposed for cancelling echo. However, they are generally not valid in all possible cases of speech transmission/reception, i.e. if only the near end subscriber talks, if only the far end subscriber talks, and if both subscribers talk (double talk situation).

For example, document WO-A-97 23055 discloses a method and device for echo cancellation using power estimation in a residual signal. Figure 1 shows an echo canceller as described in this prior art reference. "A" denotes the subscriber on the far end side of a connection and "B" denotes the subscriber on the near end side of the connection.

The input signal X(n) is attenuated by the hybrid, represented by a filter 10 with transfer function H(q^"1), where q^"1 represents the backward shift operator (q^"1X(n) = X(n-1)), and a summation unit 12, and the resulting echo signal S(n) is combined with the near end signal V(n), which may or may not contain near end speech, in summation unit 12. The attenuation of filter 10 is represented by the echo path attenuation ERL (ERL = Echo Return Loss). Thus, the resulting output signal Y(n) contains both the near end signal and echo from the far end signal. Furthermore, input signal X(n) is also forwarded to an adaptive filter 11 , which models the impulse response of the hybrid by adjusting its filter coefficients. The resulting estimate of echo signal S(n) is denoted S'(n). This estimate is, in a subtraction unit 13, subtracted from output signal Y(n) (ERLE = Echo Return Loss Enhancement represents the obtained improvement in echo attenuation), and the resulting error signal E(n) = V(n) + S(n) - S'(n) (which is referred to hereafter as the second error signal) is forwarded to adaptive filter 11 for adjustment of the filter coefficients and to the two-wire line back to far end subscriber A. The coefficients of filter 11 may be adjusted in accordance with, for example, the NLMS (Normalised Least Mean Square) algorithm.

A non-linear processor 14 receives the second error signal E(n) and outputs a processed signal ENLP(Π). An estimator 15 estimates the power of the linear error by using signals X(n) and E(n). Similarly, an estimator 16 estimates the non-linear error power by using signals E(n) and S'(n). A threshold TH(n) is computed in element 17 as a function of the outputs of the estimators 15 and 16. An element 18 computes a power estimate Re(n) in accordance with the following equation:

Re(n) = p.Re(n-1) + (1-p).E²(n) where the weighting factor p is a constant between 0 and 1 , for example 127/128. A comparator 19 compares Re(n) with TH(n), and the output signal from the comparator 19 determines the shape and attenuation of the non-linear processor 14, which filters the signal E(n) and outputs the signal E_NLp(n).

The functions of elements 14-19 can be performed by a microprocessor or a micro/signal processor combination. In case of near end subscriber talk and double talk, the proposed algorithm is unstable, because the error signal is mostly constituted by the near end subscriber talk component, which is highly decorrelated from the signal from the far end subscriber. Therefore, the adaptive filter 11 will produce wrong values and it will not be possible to fine-tune coefficient values and to reduce the residual echo.

The present invention aims at overcoming the above-mentioned drawbacks.

To this end, the present invention provides an echo cancellation device, for cancelling an echo signal output by telecommunication means, comprising: adaptive filtering means, for outputting an estimate of the echo signal, this echo cancellation device being remarkable in that: it further comprises further adaptive filtering means, receiving as an input a signal which comprises the sum of a signal originating from a first end of a connection and of the echo signal, the adaptive filtering means and the further adaptive filtering means are controlled by a first error signal, and the first error signal becomes null if, either the signal originating from the first end of the connection, or the echo signal, is suppressed by the further adaptive filtering means and if the estimate of the echo signal is equal to the echo signal.

The invention makes it possible to cancel the echo both in near end talking and double talk situations. According to a first embodiment, the first error signal is obtained as follows:

E' = AF(V + S) -V - S + S' where E' is the first error signal, V is the signal originating from the first end of the connection, S is the echo signal, S' is the estimate of the echo signal, AF is the output of the further adaptive filtering means and V + S is the input of the further adaptive filtering means. According to a second embodiment, the first error signal is obtained as follows:

E' = AF(V + S) - S' where E' is the first error signal, V is the signal originating from the first end of the connection, S is the echo signal, S' is the estimate of the echo signal, AF is the output of the further adaptive filtering means and V + S is the input of the further adaptive filtering means.

According to a third embodiment, the first error signal is obtained as follows:

E' = AF(V + S - S') where E' is the first error signal, V is the signal originating from the first end of the connection, S is the echo signal, S' is the estimate of the echo signal, AF is the output of the further adaptive filtering means and V + S - S' is the input of the further adaptive filtering means.

According to a particular feature, the echo cancellation device further comprises: double talk detection means, for detecting simultaneous speech signals originating from the first end and from a second end of the connection, and switching means, controlled by the double talk detection means, the switching means receiving as an input the first error signal and a second error signal comprising the difference between, on the one hand, the sum of the signal originating from the first end and of the echo signal and, on the other hand, the estimate of the echo, and outputting, either the second error signal, when the double talk detection means do not detect double talk, or the first error signal, when the double talk detection means detect double talk. Thanks to this switching, the invention makes it possible to have both an algorithm adapted to handle double talk situations, and an algorithm adapted to handle single line talk situations.

Other features and advantages of the present invention will appear upon reading the following detailed description of preferred embodiments, given by way of non-limiting examples.

The description refers to the accompanying drawings, in which:

- Figure 1 , already described, illustrates schematically a prior art echo canceller; - Figure 2 illustrates schematically the echo path of the PSTN echo;

- Figure 3 illustrates schematically an echo cancellation device according to the present invention, in a first particular embodiment;

- Figure 4 illustrates schematically a part of an echo cancellation device according to the present invention, in a second particular embodiment; and

- Figure 5 illustrates schematically a part of an echo cancellation device according to the present invention, in a third particular embodiment.

Figure 2 shows a Mobile Station (MS) 24 connected to the Base Station System (BSS) 26 of a cellular network. A Mobile Switching Centre (MSC) 28 is connected to an echo canceller pool 29. The arrows input in and output from the MSC 28 and the echo canceller pool 29 represent the path of the signals coming from and returned to the BSS 26 (on the left of the MSC 28) and coming from and returned to a Local Exchange (LE) of the Public Switched Telephone Network (PSTN) 31. The LE contains a hybrid unit H for converting the 4-wire line to a 2-wire line. A Stationary Telephone ST 33 is connected to the LE 31. The arrow above the MS 24 and the BSS 26 represents the path of the acoustical crosstalk and the arrow below the PSTN 31 and the ST 33 represents the path of the PSTN echo. Only minor signal disturbances occur in the echo path of the network echo canceller. However, the echo path of the echo canceller that controls acoustical crosstalk includes radio transmission. This gives rise to fundamental differences in the characteristics of the mobile echo path, in comparison with the network echo. The delay in the handset echo is long, since radio transmission requires coding and interleaving. The actual echo delay may vary, depending, by way of non-limiting examples, on the handset, system hardware, extra signal processing equipment, and the routing of the call. The characteristics of the mobile echo path are liable to vary rapidly and frequently due to changes in the position of the handset. The characteristics may also be affected by bit errors in the radio transmission, handover, or discontinuous transmission, for example. The MS echo delay can for instance vary from 120 to 320 ms. The duration of the MS echo is generally shorter than that of the PSTN echo, and its level is generally lower.

As shown in Figure 3, in an echo cancellation device according to a first embodiment of the present invention, elements 10-19 are identical to those bearing respectively the same reference numbers in the prior art device of Figure 1 and are not described here again. Thus, the telecommunication unit 10 outputs an echo signal S(n) and the adaptive filter 11 outputs an estimate S'(n) of the echo signal S(n).

The echo cancellation device according to the invention also comprises a further adaptive filter 34. The further adaptive filter 34 receives as an input a signal Y(n) which, in the particular embodiment illustrated in Figure 3, is obtained as follows: Y(n) = V(n) + S(n), where the signal V(n) originates from a first end of the connection and the signal S(n) represents the echo signal.

The adaptive filter 11 and the further adaptive filter 34 are controlled by a first error signal E'(n).

The signal Y'(n) output by the further adaptive filter 34 and the second error signal E(n) output by the subtraction unit 13 are input into a further subtraction unit 36, which outputs the first error signal E'(n) = Y'(n) - E(n).

Therefore, if the output of the further adaptive filter 34 is denoted AF(.) and if the indices "n" are omitted: E' = Y' - E = AF(Y) - E

= AF(V + S) - (Y - S') = AF(V + S) - (V + S - S') = AF(V + S) - V - S + S' = AF(V) - V + AF(S) - S + S' Thus, if the echo signal S(n) is suppressed and the signal V(n) is left unchanged by the further adaptive filter 34, the first error signal is equal to the residual error S' - S. If the echo signal estimate S'(n) is equal to the echo signal S(n), the first error signal E'(n) is therefore null.

The echo cancellation device further comprises a double talk detector unit 30, controlling a switching unit 32, to which the outputs of the subtraction units 13 and 36 are connected. Therefore, the switching unit 32 receives as an input the first error signal E'(n), output by the subtraction unit 36, and the second error signal E(n), output by the subtraction unit 13.

The double talk detector unit 30 detects simultaneous speech signals originating from the first end and from the second end of the connection (A and B in Figure 3). When the double talk detector unit 30 does not detect double talk, the switching unit 32 outputs the second error signal E(n), and when the double talk detector unit 30 detects double talk, the switching unit 32 outputs the first error signal E'(n).

Figure 4 shows a second embodiment of the echo cancellation device according to the present invention and focuses on the part thereof comprising the filter 10, the adaptive filter 11 and the further adaptive filter 34.

In this embodiment, the output of the adaptive filter 11 is connected to the input of the subtraction unit 36, in such a way that the first error signal is: E' = AF(V + S) - S' = AF(V) + AF(S) - S' Therefore, if the signal V(n) is suppressed and the echo signal S(n) is left unchanged by the further adaptive filter 34, the first error signal is equal to the residual error S - S'. If the echo signal estimate S'(n) is equal to the echo signal S(n), the first error signal E'(n) is therefore null.

Figure 5 shows a third embodiment of the echo cancellation device according to the present invention and focuses on the part thereof comprising the filter 10, the adaptive filter 11 and the further adaptive filter 34. In this embodiment, there is no subtraction unit 36, i.e. the output of the further adaptive filter is the first error signal E'(n) which is used to control both the further adaptive filter 34 and the adaptive filter 11. Moreover, the output of the subtraction unit 13 is input into the further adaptive filter 34. Consequently, the first error signal is: E' = AF(Y - S') = AF(V + S - S') = AF(V) + AF(S - S') Therefore, if the signal V(n) is suppressed and the signal S(n) - S'(n) is left unchanged by the further adaptive filter 34, the first error signal is equal to the residual error S - S'. If the echo signal estimate S'(n) is equal to the echo signal S(n), the first error signal E'(n) is therefore null.

Claims

1. An echo cancellation device, for cancelling an echo signal (S(n)) output by telecommunication means (10), comprising: adaptive filtering means (11), for outputting an estimate (S'(n)) of said echo signal (S(n)), said echo cancellation device being characterised in that: it further comprises further adaptive filtering means (34), receiving as an input a signal (Y(n)) which comprises the sum of a signal (V(n)) originating from a first end of a connection and of said echo signal (S(n)), said adaptive filtering means (11) and said further adaptive filtering means (34) are controlled by a first error signal (E'(n)), and said first error signal (E'(n)) becomes null if, either said signal (V(n)) originating from said first end of said connection, or said echo signal (S(n)), is suppressed by said further adaptive filtering means (34) and if said estimate (S'(n)) of said echo signal (S(n)) is equal to said echo signal (S(n)).

2. An echo cancellation device according claim 1 , characterised in that said first error signal (E'(n)) is obtained as follows:

E' = AF(V + S) - V - S + S' where E' is said first error signal, V is said signal originating from the first end of said connection, S is said echo signal, S' is said estimate of said echo signal, AF is the output of said further adaptive filtering means (34) and V + S is the input of said further adaptive filtering means (34).

3. An echo cancellation device according to claim 1 , characterised in that said first error signal (E'(n)) is obtained as follows:

E' = AF(V + S) - S' where E' is said first error signal, V is said signal originating from the first end of said connection, S is said echo signal, S' is said estimate of said echo signal, AF is the output of said further adaptive filtering means (34) and V + S is the input of said further adaptive filtering means (34).

4. An echo cancellation device according to claim 1 , characterised in that said first error signal (E'(n)) is obtained as follows: E' = AF(V + S - S') where E¹ is said first error signal, V is said signal originating from the first end of said connection, S is said echo signal, S' is said estimate of said echo signal, AF is the output of said further adaptive filtering means (34) and V + S - S' is the input of said further adaptive filtering means (34).

5. An echo cancellation device according to any of the preceding claims, characterised in that it further comprises: double talk detection means (30), for detecting simultaneous speech signals originating from said first end and from a second end of said connection, and switching means (32), controlled by said double talk detection means, said switching means receiving as an input said first error signal (E'(n)) and a second error signal (E(n)) comprising the difference between, on the one hand, the sum of said signal (V(n)) originating from said first end and of said echo signal (S(n)) and, on the other hand, said estimate (S'(n)) of said echo, and outputting, either said second error signal (E(n)), when said double talk detection means do not detect double talk, or said first error signal (E'(n)), when said double talk detection means detect double talk.