Signal idealisation in phonetic models

The following properties of speech signals are generally excluded from phonetic models, though the effects may be strong, and there is considerable research being conduced on them (see the Figure):

1. Amplitude effects.
   The perceived amplitude of a signal depends not only on its amplitude at the source, but also on the distance of the observer from the source.
2. Phase effects.
   Signals are located in space, and their position can be located perceptually - the `stereo' effect - by means of their phase differences.
3. Frequency effects.
   Signal sources move in space, leading to an apparent change in frequency in proportion to the relative speed towards or away from the observer - the Doppler effect. The sounds of moving cars, sirens, etc. sound higher when the vehicle is approaching, only to change when the vehicle goes past, and fall in perceived pitch when the vehicle is moving away.
4. Perceptual effects on speech production.
   Speech signal production is affected by perception of the environment and by feedback from the signal, via the environment, as it is produced. Two of these effects are the following:
   1. The Lombard effect.
      The perceptual environment of the speaker the production of speech. The term `Lombard' alludes to the particularly loud environment of the London Stock Exchange, Lombard Street, in the City of London, and the specific style of speech production of the brokers in this environment. Speech in loud vehicles, at football matches, or in school playgrounds could also be cited. Effects at the quiet end of the scale will be familiar from libraries, churches, museums, bedrooms.
   2. The Lee effect.
      When speech is played back to the speaker with a delay of about 300 msec (approximately the length of one syllable), a form of stuttering is triggered.
   3. Pathology.
      Judgments of `normal speech' underlie the mainstream of phonetic modelling of speech; serious variation due to accident or illness are dealt with in the specialisation of Clinical Phonetics, and other variations such as the influence of smoking or alcohol have also been researched.
   4. Standard speech.
      The speech of a socially well-established group, such as speakers of a standard language, is the main object of phonetic modelling; the phonetics of dialect and sociolect variation is a specialisation which was, in fact, historically the first motivation for establishing the science of phonetics in the 19th century.
   5. Clear speech.
      The pronunciation of words in isolation, the reading aloud of text, have traditionally been the objects of phonetic transcription and experimentation. More recently, the phonetics of continuous speech and spontaneous dialogue have received increasing attention in the context of the development of advanced speech processing systems.