T 2271/18 (Acoustic-feedback reduction/SIVANTOS) 25-03-2021

1. The present application

The present invention concerns feedback suppression by means of an adaptive filter in an audio device. The

adaptive filter is controlled via a step-size parameter my, which is described in the present application in the context of the well-known normalised least mean square (NLMS) algorithm. Feedback in the audio device is eliminated completely by setting the step-size parameter to a high value such as "0.06". Under these high step-size parameter conditions, an input signal from the environment which is strongly correlated can, in the configuration of the invention, still lead to an erroneous detection of feedback. If such an erroneous detection occurs, the high value of the step-size parameter is reduced in order to avoid artefacts.

1. Main request: claim 1 - features

Claim 1 of the main request comprises the following limiting features (with the board's labelling):

(a) A method for reducing acoustic feedback of an audio device by a feedback compensator, said feedback compensator having at least one adaptive filter,

comprising the steps of:

(b) setting a step-size parameter my of said adaptive filter at a high level to eliminate acoustic feedback;

(c) decreasing said parameter my when a critical excitation signal is detected.

2. Main request: claim 1 - clarity

Claim 1 of the main request is unclear, for the following reasons:

2.1 As regards feature (a), the structural relationship between the "feedback compensator" and the "audio device" is unclear. In fact, the claimed "feedback compensator" represents a "black box", specifying only an input parameter and the result that should be achieved, rather than being specified in terms of structural features actually solving the technical problem posed in the passage on page 6, lines 7 to 9 of the description as filed ("subjective problem"), i.e. "combining the reduction or even elimination of acoustic feedback reduction with an efficient avoidance of artifacts". The board refers in this respect to T 2065/10 (Reasons 2.2.1) and notes in particular that the name given to a particular black box, which in this case may express the aim of compensating feedback, does not provide details on how the black box operates, in the current case on how feedback is compensated.

2.2 As regards feature (b), the technical meaning of the "step-size parameter my" and the so-called "high level" in the context of reduction or elimination of acoustic feedback is entirely unclear. While the skilled reader may be aware, based on their common general knowledge, of the gain being a critical parameter in relation to feedback and the adaptation rate of an adaptive filter being important to strike a balance between a quick response and a stable adaptation, the same is not valid for the size of a step which relates in an unspecified way to the adaptive filter of a general feedback compensator. Moreover, the appellant's allegation that the term "step-size parameter of an adaptive filter" would be readily understood by the skilled reader, in particular in connection with the problem of feedback and the explicit denomination of the step-size parameter as "my" is not convincing. This is because it belongs to the skilled reader's common general knowledge that adaptive filters require the solution of a computational optimisation problem involving a series of steps to choose the parameters of their transfer function appropriately. While each step in this series is typically characterised by a particular parameter, which may be used to define an adaptation rate (e.g. a rate of convergence "my" or the change of an error tolerance "epsilon" over the iterations), it is by no means implicit that this parameter always involves a particular step size. Hence, the "step-size parameter" is not inherently defined by the general "adaptive filter". Furthermore, as regards the term "high level", it is unclear to a skilled reader which numerical value to attribute to that relative term for setting this parameter as required by feature (b). The argument that the skilled reader would simply choose a "sufficiently large value" that efficiently prevents feedback is not persuasive because feature (c) requires that the step size be reduced again: the step-size parameter must therefore be set higher than this "sufficiently large value", leaving the skilled reader in doubt as to which value would be high enough.

2.3 As regards feature (c), the term "critical excitation signal" is unclear because claim 1 does not allow the skilled reader to distinguish such an arbitrary "critical signal" from "uncritical" trigger signals in a meaningful way.

2.4 In conclusion, claim 1 of the main request does not comply with Article 84 EPC.

3. First to fourth auxiliary requests: admittance

3.1 The first to fourth auxiliary requests were filed after notification of the summons to oral proceedings (cf. point III above). In accordance with Article 13(2) RPBA 2020, such an amendment to the party's appeal case shall, in principle, not be taken into account unless there are exceptional circumstances, which have been justified with cogent reasons by the party concerned. In the application of the exercise under Article 13(2) RPBA 2020, the criteria mentioned in Article 13(1) RPBA 2020 may be used. One such criterion concerns whether the party has demonstrated that the amendments to its appeal case, prima facie, overcome the issues raised by the board and do not give rise to new objections.

3.2 The appellant justified the amendments underlying the first to fourth auxiliary requests by

(i) referring to the introduction, in the board's communication pursuant to Article 15(1) RPBA 2020, of a lack of clarity as a new objection in the appeal proceedings;

(ii) highlighting that the newly found lack of clarity was motivated by a list of multiple clarity objections;

(iii) indicating that the board had implied that the list of clarity objections was not exhaustive.

3.3 While arguments (i) to (iii) may in principle constitute "cogent reasons" (cf. the third paragraph of the explanatory remarks to Article 13(2) RPBA 2020), it is within the board's power to raise new objections during the appeal proceedings (cf. G 10/93; T 862/16, Reasons 8.3.1). In particular, in order to help concentration on essentials during the oral proceedings, a board's communication pursuant to Article 15(1) RPBA 2020 is based on a thorough analysis of the case and is meant to draw attention to matters that may be of particular significance for the decision to be taken, while the board "may" also provide a preliminary opinion (cf. Article 15(1), fifth sentence, RPBA 2020). The level of detail with which this analysis is reflected in the communication, in particular the extent to which a preliminary opinion is provided, is, however, up to the particular board's discretion, in view of the circumstances of the case. Accordingly, even a high level of detail provided in a communication issued under Article 15(1) RPBA 2020 can by no means be interpreted as an "invitation" for a party to amend its appeal case by filing amended claims (cf. for instance T 1459/11, Reasons 3.2; T 752/16, Reasons 3.4).

In other words, a clear and detailed preliminary opinion provided by a board - rather than merely "drawing attention to matters that seem to be of particular significance for the decision to be taken" (cf. Article 15(1), fourth sentence, RPBA 2020) - is predominantly intended to give the party(ies) an opportunity to thoroughly prepare their arguments in response to it but not to file new submissions, such as new sets of claims, and to thereby arguably shift the focus on the issues on file to be decided in appeal proceedings. In particular, amendments submitted in response to such a preliminary opinion cannot give rise to "exceptional circumstances" within the meaning of Article 13(2) RPBA 2020.

3.4 Even if it was accepted that there were cogent reasons justifying the presence of "exceptional circumstances" here, the board concludes that the above auxiliary requests are not clearly allowable for the following reasons:

3.4.1 None of the four versions of claim 1 of those auxiliary requests overcomes all of the objections raised in point 2 above as to claim 1 of the main request. For instance, all four versions of claim 1 attempt to overcome the issue of the poorly defined step-size parameter (cf. point 2.2 above) solely by the insertion of the phrase "wherein the step-size parameter my controls the adaptation speed of the adaptive filter". However, not all adaptive filters rely on an iterative scheme involving one or more steps that are characterised by a (variable) step size. Accordingly, not all adaptive filters allow to define a particular adaptation speed, which must, due to its implied movement, relate to a particular step-size change over time. Rather, iterative schemes are often characterised by a more general adaptation rate. The inserted phrase therefore does not, prima facie, resolve the issue with respect to the poorly defined step-size parameter.

3.4.2 Furthermore, the amendments underlying the versions of claim 1 of the first to fourth auxiliary requests give even rise to new objections. As a mere example, the phrase "in this case using the feedback detector (16) to detect a critical excitation signal" (emphasis added), which is present in all of these four versions of claim 1, would confuse the skilled reader in that it suggests to use the feedback detector only in particular cases, whereas the feedback detector in the circuit of Figure 3 as originally filed is manifestly always active. This is because the circuit does not involve any switch designed to take the feedback detector out of the processing path under predetermined conditions. This phrase therefore further aggravates the lack of clarity of claim 1 of the main request.

3.5 Hence, the first to fourth auxiliary requests were not taken into account in these appeal proceedings (Article 13(2) RPBA 2020).

4. Fifth to ninth auxiliary requests: claim 1 - clarity

4.1 None of the five versions of claim 1 in accordance with the fifth to ninth auxiliary requests remedies the clarity deficiencies of claim 1 of the main request.

In particular, the version of claim 1 of the ninth auxiliary request is the only one of these five versions that tries to address the issue of the poorly defined step-size parameter (cf. point 2.2 above), namely by the insertion of a phrase which is, apart from the term "step-size parameter" being preceded by the word "said" instead of by the definite article, identical to the one mentioned in point 3.4.1 above for claim 1 of the first to fourth auxiliary requests. For the reasons set out in point 3.4.1 above, this phrase does, consequently, not resolve the clarity issue of point 2.2 above.

4.2 As a consequence, claim 1 of the fifth to ninth auxiliary requests does not comply with Article 84 EPC either.

5. Tenth auxiliary request: admittance

5.1 The tenth auxiliary request is identical to the second auxiliary request underlying the decision under appeal, which was not admitted into the proceedings by the examining division.

5.2 It is generally accepted that, for the purpose of reviewing first-instance discretionary decisions, it is not for the board to review all the facts and circumstances of a case as if it were in the place of the department of first instance and decide whether or not it would have exercised discretion in the same way.

5.3 In the present case, there are no indications that the examining division did not exercise their discretion under Rule 137(3) EPC in a proper manner and following the right principles. The appellant did not provide any arguments to contest this. Rather, only arguments in support of compliance with Article 123(2) EPC were submitted.

5.4 Moreover, the board sees no other reasons to admit the tenth auxiliary request into the proceedings.

5.5 As a result, the board does not consider it justified to overrule the discretionary decision of the examining division, and to exercise its own discretion under Article 12(4) RPBA 2007 (cf. Article 25(2) RPBA 2020) to admit the tenth auxiliary request into the appeal proceedings.