Design and evaluation of tinnitus synthesis methods: From spectral to spatial matching

Purpose
This study was designed to investigate methods to help patients suffering from unilateral tinnitus synthesizing an auditory replica of their tinnitus.

Materials and methods
Two semi-automatic methods (A and B) derived from the auditory threshold of the patient and a method (C) combining a pure tone and a narrow band-pass noise centred on an adjustable frequency were devised and rated on their likeness over two test sessions. A third test evaluated the stability over time of the synthesized tinnitus replica built with method C, and its proneness to merge with the patient's tinnitus. Patients were then asked to try and control the lateralisation of this single percept through the adjustment of the tinnitus replica level.

Results
The first two tests showed that seven out of ten patients chose the tinnitus replica built with method C as their preferred one. The third test, performed on twelve patients, revealed pitch tuning was rather stable over a week interval. It showed that eight patients were able to consistently match the central frequency of the synthesized tinnitus (presented to the contralateral ear) to their own tinnitus, which leaded to a unique tinnitus percept. The lateralisation displacement was consistent across patients and revealed an average range of 29dB to obtain a full lateral shift from the ipsilateral to the contralateral side.

Conclusions
Although spectrally simpler than the semi-automatic methods, method C could replicate patients' tinnitus, to some extent. When a unique percept between synthesized tinnitus and patients' tinnitus arose, lateralisation of this percept was achieved.

Statistically-constrained shallow text marking:Techniques, evaluation paradigm and results

We present three natural language marking strategies based on fast and reliable shallow parsing techniques, and on widely available lexical resources: lexical substitution, adjective conjunction swaps, and relativiser switching. We test these techniques on a random sample of the British National Corpus. Individual candidate marks are checked for goodness of structural and semantic fit, using both lexical resources, and the web as a corpus. A representative sample of marks is given to 25 human judges to evaluate for acceptability and preservation of meaning. This establishes a correlation between corpus based felicity measures and perceived quality, and makes qualified predictions. Grammatical acceptability correlates with our automatic measure strongly (Pearson's r = 0.795, p = 0.001), allowing us to account for about two thirds of variability in human judgements. A moderate but statistically insignificant (Pearson's r = 0.422, p = 0.356) correlation is found with judgements of meaning preservation, indicating that the contextual window of five content words used for our automatic measure may need to be extended. © 2007 SPIE-IS&T.