Instrumentation & Measurement Magazine 24-6 - 13

Measuring Cognitive Status
from Speech in a Smart Home
Environment
Kathleen C. Fraser and Majid Komeili
T
he population is aging and becoming more techsavvy.
The United Nations predicts that by 2050, one
in six people in the world will be over age 65 (up from
one in 11 in 2019), and this increases to one in four in Europe
and North America. Meanwhile, the proportion of American
adults over 65 who own a smart phone has risen 24 percentage
points from 2013-2017, and the majority have Internet in their
homes [1]. Smart devices and smart home technology have
profound potential to transform how people age, their ability
to live independently in later years, and their interactions with
their circle of care.
Cognitive health is a key component to independence and
well-being in old age, and smart homes present many opportunities
to measure cognitive status in a continuous, unobtrusive
manner. In this article, we focus on speech as a measurement
instrument for cognitive health. Existing methods of cognitive
assessment suffer from a number of limitations that could
be addressed through smart home speech sensing technologies.
We begin with a brief tutorial on measuring cognitive
status from speech, including some pointers to useful opensource
software toolboxes for the interested reader. We then
present an overview of the preliminary results from pilot studies
on active and passive smart home speech sensing for the
measurement of cognitive health, and conclude with some recommendations
and challenge statements for the next wave of
work in this area, to help overcome both technical and ethical
barriers to success.
Measuring Cognitive Decline from
Speech
Declines in cognitive ability can occur as a result of a number of
underlying physical factors: traumatic brain injury, stroke, and
neurodegenerative diseases such as Alzheimer's disease (AD),
among others. The set of cognitive symptoms associated with
neurodegenerative diseases is known as dementia. Since dementia
develops slowly over time, there is increasing interest
in methods to detect it at the earliest possible stages, when interventions
that aim to slow down the decline process are most
effective. It is estimated that approximately half of dementia
September 2021
cases go undiagnosed, which can lead to poorer outcomes for
both the individual and their caregivers [2].
Part of the challenge lies in the fact that 'cognition' is not
a physical property to be measured, and all measures of cognition
are by necessity indirect. A physician may look at an
MRI image and see the patterns of atrophy in the brain, but
in many cases such information is only roughly correlated
with the actual behavior and function of the patient, particularly
at the early stages. Typically, cognition is measured
through standardized 'cognitive tests' which target different
areas of cognition and are validated against existing population
norms. However, as a measurement instrument, these
tests suffer from both practical and theoretical limitations: they
are administered in a clinic by a trained professional, which
is expensive and time-consuming, and also affects the measurement
itself, since social anxiety and pressure to do well
can potentially lead to poorer performance. Furthermore, the
measurements are not generally repeatable, due to a learning
effect, making measurement reliability difficult to estimate. At
the same time, repeated measurements are highly desirable, as
it is known that people with dementia have " good days and
bad days, " and cognitive ability can vary from day-to-day. Alternative
measurement instruments are needed, particularly
those that can be applied outside of the clinic.
Speech and language offer a natural and incredibly rich
view into cognitive function. In dementia, various cognitive
areas such as memory, executive function, attention, processing
speed, and language itself may be affected, and these
deficits can all show up in what a person says and how they say
it. Therefore, there has been growing interest in automatically
detecting the changes in speech and language that indicate underlying
cognitive decline.
Preliminary lab-based studies have supported the hypothesis
that speech carries information about cognitive status.
For example, a machine learning classifier trained on linguistic
and acoustic features extracted from short speech samples
was able to distinguish between participants with and without
AD with 81% accuracy [3]. Other work used automatic conversation
analysis to differentiate between participants with
IEEE Instrumentation & Measurement Magazine
1094-6969/21/$25.00©2021IEEE
13
Instrumentation & Measurement Magazine 24-6

Table of Contents for the Digital Edition of Instrumentation & Measurement Magazine 24-6
