Articulating the Articulation Index

February 18, 2009 by
Filed under: Voice and Sound Quality Testing Software 

In the next couple of posts we will present existing approaches to measure voice quality automatically. This is an important issue as for telecom industry as well as for media industries and at a certain point we’ll present a proof for that.

One of the methods is the so-called Articulation Index (AI). The idea is that the whole frequency range of speech signal is divided into 20 bands and the signal/noise ratio is determined within the band. The band broad is defined in such a way, that every band contributes equally in speech perception. The signal/noise ratio is calculated within every band. Articulation index is supposed to be equal the weighted total of the band values.

And one of the main disadvantages of the method is that articulation index does not take into account the properties of hearing and speech production, although it directs towards speech signal.

Share

Tags:

Comments

2 Comments on Articulating the Articulation Index

  1. Christine Rankovic on Mon, 13th Jul 2009 13:31

    2. The disadvantage of the articulation index that you cite is incorrect. The articulation index (AI) was designed after careful consideration of the properties of hearing and speech. The AI is based on 30 years of extensive research at AT&T Bell Telephone Laboratories led by Harvey Fletcher. Properties of hearing considered include tone and speech detection thresholds, speech loudness growth with intensity, including consideration of speech loudness limits which–when exceeded–reduce intelligiblity, critical bands of hearing (frequency resolution capability of the ear), masking effects (simultaneous and forward masking) caused both by noise and by speech masking itself, and much more.

    For speech production, the range of speech both in frequency and time was measured and built into the AI calculation, as was the distribution of short-term levels of speech. The contribution of different frequency regions to intelligibility was derived from filtered speech experiments using highly trained listeners and talkers. Calibration was studied carefully, and original measurement instruments were built just for the AI work. They also studied aspects of speech waveforms (frequency content and intensity) for individual speech sounds in order to associate their characteristicss with intelligibility.

    The effort to develop the articulation index was much greater than most people realize. See Articulation Index History on my website (www.articulation.com) for a short timeline.

  2. admin on Tue, 14th Jul 2009 01:40

    3. We do appreciate the input of the Articulation Index to research in the field of voice quality and a 30 years experience is a quite impressive research period. However, our software is targeted at real life application minimizing human interaction and increasing objectiveness of the quality estimation. Perhaps we don’t have a 30 year experience developing our methods and software, but do take into account previous research and modern scientific achievements, and we emphasise that one can easily download our software and do testing by himself finding the proof to the advantages of our method we state.

Tell me what you're thinking...
and oh, if you want a pic to show with your comment, go get a gravatar!