Another challenge affecting EEG use is the high cost associated with such systems. However, a very high demand for neurogaming is driving down the cost of EEG headsets. Neurogaming applications allow users to interact with the gaming software and the software makes adaptations based on the EEG data collected from the users. An example of a single channel headset is shown in Figure 2. This headset costs approximately $100, and research capability can be added to the system for an additional fee.
Auditory stimuli also evoke magnetic fields which can be recorded using magnetoencephalography (MEG) techniques. The current devices used for recording MEG activity are relatively expensive, large, and thus not portable. However, miniaturized MEG devices are being developed that can be as small as a nano chip to allow discrete and easy collection of MEG data.
AVAILABILITY OF SOPHISTICATED TECHNIQUES FOR ANALYZING NEURAL DATA
At present, there are no well-established and sophisticated clinical techniques that enable objective determination of WRS. However, other analytical approaches may be applied such as those used by investigators to analyze neural data evoked by presenting words to the listeners or by asking listeners to imagine the pronunciation of specific words. There are some similarities in the neural activation when a participants is listening to spoken words or imagining the pronunciation of the same words; in both cases the auditory cortex is activated.
An important motivation in developing sophisticated neural pattern analyses techniques is to allow patients who are unable to express their thoughts (e.g., locked-in-syndrome) to communicate through computers. For example, Torres-García et al. focused on the recognition of neural patterns evoked by internally pronounced Spanish words corresponding to English words such as up, down, left, right, etc (Expert Syst Appl 2016;59:1 http://ow.ly/yEH0300TlDE). The EEG data was obtained by placing 14 electrodes on the scalp. Using sophisticated machine learning techniques, the investigators were able to show high accuracy (average 70%, maximum accuracy 91.47%) in decoding the imagined words based on the neural data evoked by the imagined words.
Neural pattern analyses techniques should also be developed to see if, instead of using a typed password, users could log onto their electronic devices using “pass-thoughts.” Such “pass-thoughts” are resistant to other individuals stealing the passwords through shoulder-surfing or using a password hacking software. In another study, participants were asked to imagine that they were singing a song or reciting a passage for 10 seconds, along with several other tasks such as listening to a tone (Adams. Financial Cryptography and Data Security. Berlin: Springer, 2013). Neural data was collected using a single channel EEG headset while the participants were performing such tasks. Using sophisticated analytical techniques, the investigators were able to achieve 99 percent user authentication accuracy. These current technological developments indeed show the future possibilities of objective WRS assessment.
Potential Steps in the Development of an Objective WRS Procedure:
* Select Standardized Word Lists.
* Record the words and all foils using a single speaker.
* Record neural data evoked for all recorded words, including foils using participants with excellent WRS.
* Develop normative database of specific neural patterns associated with each word, including all foils, by submitting the normal neural data to sophisticated analytical techniques (e.g., machine learning).
Potential Steps in Objectively Assessing Speech Recognition Scores
* Present the recorded words to the patient.
* Record the neural data evoked by each word.
* Compare the neural data evoked by each word to the normal expected neural patterns from the previously established database, related to that word.
* If there is a match between the patient's neural activity and the expected neural pattern associated with that word, the word can be assumed to be correctly encoded with the patient's neural system.
* If there is a significant mismatch, inaccurate word recognition can be expected. As an advanced step, the inaccurate neural pattern could be compared to the neural pattern evoked by the foils of the words. For example, if the presented word was “fall” and the neural pattern matches to the foil “ball”, then the misperception can be predicted as being “ball” instead of “fall.”
Copyright © 2016 Wolters Kluwer Health, Inc. All rights reserved.