Using Forward Masking Patterns to Predict Imperceptible Information in Speech for Cochlear Implant Subjects
Repository Usage Stats
Forward masking is a phenomenon that occurs when one stimulus (masker) elevates the threshold for perception of a subsequent stimulus (probe). In cochlear implant listeners, it has been hypothesized that interactions between electrodes (termed channel interactions) can have a deleterious effect on speech recognition (Chatterjee and Shannon, 1998; Throckmorton and Collins, 1999). Forward masking is one measure that has been used to assess channel interactions (Shannon, 1983; Tong and Clark, 1986; Lim et al., 1986; Shannon, 1990; Blamey and Dooley, 1993; Chatterjee and Shannon, 1998), and it has also been proposed as a mechanism that could be used to assess information that is being presented to but is not received by the user (Throckmorton and Collins, 1999; Nogueira et al., 2005). Determining information that is not received by the user, or is masked, has the potential to provide guidance for the design of new speech processing algorithms that either work to reduce lost information or substitute alternate information.
Nogueira et al., (2005) investigated the potential for using forward masking to estimate masked information and found a limited benefit with information substitution; however, their findings were based on normal hearing psychoacoustic forward masking patterns (Nogueira et al., 2005). Given that psychophysically measured forward masking patterns vary from subject to subject and electrode to electrode (e.g. Shannon, 1983; Shannon, 1990; Chatterjee and Shannon, 1998), measuring subject- and electrode-specific patterns has the potential to provide a more accurate assessment of masked information. In addition, the experimental time required to gather the full set of psychophysical forward masking patterns for all electrodes is far too long for clinical relevance (Throckmorton and Collins, 1999). Recently, it has been suggested that forward masking patterns can be measured physiologically via the electrically evoked compound action potential (ECAP) (e.g. Hughes and Stille, 2009) which may make it possible to measure the information necessary for a subject-specific assessment of masked information in a clinically relevant time frame.
This study utilized ECAP measurements to estimate the forward masking patterns in subjects, and these masking patterns were used to estimate the masked stimuli in speech. The estimates were validated using a speech reception threshold task to assess whether speech recognition is affected by removing "masked" pulses from a subject's stimulation pattern. The results of this study suggest that forward masking patterns measured via ECAPs could potentially determine information that is not being perceived by the user.
The masked stimuli were also used to statistically evaluate the segments of speech that are most vulnerable to masking. These results suggest that the amount of masking that occurs per phoneme depends on the characteristics of the subject-specific masking patterns.
DepartmentElectrical and Computer Engineering
More InfoShow full item record
This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 3.0 United States License.
Rights for Collection: Masters Theses