GVC applications in Neurology can measure the severity of some diseases by measuring aspects of speech or other biological signals. Applications that we are currently focusing on measure the quality of the voice: the user speaks for a few seconds, and the GVC Voice Disorder Detection algorithm measures hundreds of acoustic properties of the user’s voice and distills from these cues an assessment of the user’s voice problems, and thereby of the degree to which the disease impacts the quality of the voice.
As with GVC Emotion Recognition, the acoustic cues that we measure are numerous: static and dynamic properties of pitch, intensity, resonances, dullness, sharpness, softness, tempo, phrasing, and many more. Existing voice disorder software typically reports voice-quality-related acoustic properties such as jitter, shimmer, and harmonicity. The GVC Voice Disorder Detection algorithm measures these as well, but as some of the inputs to the voice quality mapping algorithm, not as its outputs.
We use clinical trials to teach the GVC Voice Disorder Detection software to assess the acoustic cues as indicators of voice quality. Patients read some text, or count to twenty, or answer simple questions asked by the practitioner. The physician also supplies us with the results of an independent assessment of the stage of the disease. When combining these sources of information, GVC’s algorithm optimizes its knowledge of the association between the acoustic cues and the severity of the disease as judged by experts.
The software becomes a diagnostic device whose report correlates with the severity of the disease, especially with the degree to which the disease has managed to affect voice quality. In some cases, the patient may be able to do measurements at home, perhaps by calling the practitioner though the phone, without having to come to the hospital.
The software can support medication. The symptoms of some if the diseases that affect voice quality (Parkinson’s disease, for instance) are ameliorated by medication. To the extent to which this medication improves voice quality as well, the GVC Voice Disorder Detection software can help in assessing whether the correct dose has been applied, which is a major issue in the treatment of some of these diseases.
A future development could be the early detection of some diseases. In cases where the GVC Voice Disorder Detection is particularly sensitive, it might be thinkable that diseases can de detected at such an early stage that their full expression can still be prevented. Small decelerations or synchronization issues in articulation may not be noticed by human listeners, but might be detectable by an algorithm that looks for them.