Análisis mutacional de los pacientes AF

The Dolby Digital 5.1 surround sound format may in itself bring advantages for hearing impaired and other television viewers. Some research suggests that there may be some benefits for television sound by the addition of a central loudspeaker, as is used in 5.1 surround sound systems, compared to a central ‘phantom’ stereo image.

Often, where both 5.1 surround sound and two channel stereo broadcasts take place only one mix is carried out in 5.1 and an automated down mix used for stereo broadcast. Increasingly though, the 5.1 mix is the only available broadcasted format and

downmixing occurs at the set top box in the users’ home. It is suggested by Dressler (Dressler.R., 1996) that the downmix process, whereby a 5.1 surround sound audio stream is converted for 2 channel playback, may distort the mix in such a way as to reduce intelligibility by altering “the subjective balance of the mix”. Holman (1991) suggested that the addition of a central loudspeaker made the material easier to

understand although stated that this may not actually produce greater intelligibility. This effect, leading to an apparent difficulty in understanding, is a result of acoustical

crosstalk (Holman.T., 1996) that occurs when two identical signals arrive at the ear with one slightly delayed compared to the other. This produces a comb filtering effect that cancels out some frequencies in the audio.

Additionally the comb filtering effect has been found to be detrimental to the listening experience more generally. Commenting on frequency response problems caused by signal path delays David Clark states that “Clearly the ‘phantom’ center is an entirely different listening experience than pure left or pure right. One might ask if stereo is deeply flawed as [a] sound reproduction technique or if interference notches should simply be ignored” (Clark, 1983 cited in Vickers, 2009a). Impacts for listeners such as these that go beyond intelligibility mean that considerable efforts have gone into attempts to remove, or reduce the impact of crosstalk. Methods have been proposed to reduce the impact of this crosstalk by Cooper and Bauck (1989) and Bauck and Cooper

(Bauck and Cooper, 1992) but these may be impractical in the context of television viewing as they utilise crosstalk cancellation techniques that rely heavily on the listener being in the ideal listening position. Clearly in a home environment this is very rarely the case. Vickers (2009a) recognises this and goes further pointing out that “when the listener is not equidistant from the speakers, the comb filter peaks and nulls will be in different frequency locations at the two ears”. The resultant comb filtered perception in any given location in the room then becomes very unpredictable and impossible to compensate with additional comb filters. There is some debate as to the specific cause of intelligibility problems resulting from crosstalk. Bucklein (1981) suggests that intelligibility difficulties may actually be made worse by peaks resulting from the crosstalk effect, rather than the troughs, as might be assumed however the underlying problem remains regardless of which effect of crosstalk is most detrimental. Other approaches to reduce crosstalk impact have been suggested; decorrelation methods (Kendall, 1995; Boueri and Kyriakakis, 2004) have been suggested so as to randomise the effects of crosstalk and so make the effects less prominent, this can be seen as a signal processing equivalent of relying on room reflections to even out responses, but others have found artefacts and distortions from these methods which, with musical content, have manifested themselves as unacceptable timbre change (Augspurger et al., 1989).

Vickers also suggests a further possibility for defeating crosstalk; by deriving a centre channel from two channel stereo content which would then be presented as a real, rather than a phantom, source. He suggests a method for accomplishing this using frequency domain upmixing (Vickers, 2009c) and provides a useful review of upmixing methods (Vickers, 2009b). Clearly this would be a useful direction if it was effective as side channels (L and R) could be reduced with reference to the new centre channel content in order to improve intelligibility for people with a hearing impairment. His research suggested that existing upmixing algorithms either provided inadequate centre channel separation or produced ‘watery sound’ or ‘musical noise’ artefacts (Vickers, 2009b) although formal subjective testing was not applied to assess this thoroughly (Vickers, 2013). These methods are specifically about spatial decomposition, rather than signal

separation, a different approach which is beyond the scope of this thesis. Goodwin and Jot (2007) make reference to primary-ambient decomposition in extracting ambient information from stereo signals using principal component analysis and a variation on principal component analysis for accessible TV audio is implemented in chapter 6 of this thesis.

Much of the limited literature around the subject of broadcast audio for hearing impaired people covers signal processing methods but is speculative as regards the impact on people, particularly hearing impaired people. There is a substantial gap in the literature of robust subjective assessment of how such processes affect people with hearing impairments and this thesis aims to fill some of these gaps.

When carrying out subjective assessments for perceptual aspects such as clarity there is an issue of the degree that visual cues can influence understanding of test material, in addition to the intelligibility of the audio information. Grant at al (1998) found great variability between participants in their ability to utilise audio/visual integration to improve understanding of material but estimated potential improvements using visual content of up to 26% in some individuals. Early research by Sumby and Pollack (1954) indicates that the visual contribution to speech intelligibility increases significantly as speech to noise ratio decreases. In the VISTA project (Carmichael et al., 2003) a high degree of ‘speech reading’ was recognised as being attempted by older participants in attempting to understand an avatar with a synthetic voice, this was partially

unsuccessful owing to lip sync problems although this in itself indicates a degree of reliance on visual cues for older users. Other research (Beerends and De Caluwe, 1999) shows biasing in assessments of AV media quality from both audio and visual

interactions. The research indicates that, in their study, quality of visual presentation had more impact on assessments of audio quality than quality of audio presentation had on assessments of visual quality. In each case significant influence was demonstrated. For the audio researcher this is potentially problematic and care must be taken to ensure that video quality is consistent throughout AV media presentation. Audio quality

tests however, for example where audio quality may not be the only descriptor under scrutiny, audio-visual presentation will be necessary andany test procedures

incorporating visual material must also be carefully designed to eliminate any bias resulting from visual cues in the media.