Can Disorders be Heard? With Henry O’Connell

Henry O’Connell, Founder and CEO of Canary Speech joins guest-host Erica Robinson from Solace LifeSciences, Inc. to talk about how the Canary Speech app uses AI and machine learning to provide 200 thousand data points to analyze stress levels in speech patterns.

 

“When you’re looking at 200,000 data points, you can be more accurate, you can be more specific. We can tell the severity of someone’s stress in the morning and after doing a meditation or an event where they’re calming themselves. After they go walking or after they exercise, they can measure again and see the impact of that in their lives.”

— Henry O’Connell

 

Ever thought three minutes of light conversation could give you that much information?

Tune in and get ready to take notes.

 

Listen to This Is NuCalm on Apple & Spotify!

 

Henry O’Connell has over 20 years of executive and C-level experience. Following graduate school, O’Connell began his career at the National Institutes of Health in a neurological disease group and continued on to a successful business career specializing in turnaround situations in the tech industry.

 



Key Takeaways

[1:00] David welcomes Henry O’Connell as well as today’s guest host, Erica Robinson. Erica starts off the conversation with a question on the origins of Canary Speech.

 

[4:00] What are the underlying speech markers of the human condition? Henry speaks to the features that the Canary Speech app provides.

 

[5:00] There are multiple speech recognition software apps available, however, Henry offers a few critical reasons why their app differs and how it makes it much more performant — regardless of the spoken language.

 

[7:53] AI training for the Canary Speech algorithm was enabled through a multitude of clinical and pharma studies and resulted in models that are aligned with industry standards. Henry shares some of the research they did as well as the results they generated

 

[10:00] Whether for your own baseline or as a community comparison tool, Canary Speech’s scores can serve as a tool to monitor your overall sympathetic health.

 

[11:13] Henry talks about Canary Speech performance as compared to the usual methods (cortisol, saliva testing, HRV, etc.)

 

[14:16] What studies are currently being done in the mental health field with Canary Speech? Henry shares how they are trying to identify mental illness comorbidities as vectors of health improvement.

 

[19:35] Henry shares some personal examples of using the app and how it has impacted his own life experience.

 

[23:15] Canary Speech’s newest projects and opportunities!

 

[27:17] Henry shares what the last book he read was: check out Glenn Beck.

 

[28:38] Erica thanks Henry for sharing his knowledge and app with the mastermind and invites listeners to keep an ear out for a Q&A episode with Henry in the future.

 

Continue on your journey and until next time, breathe deep, relax, and keep looking forward.

 

Mentioned in this episode

NuCalm

Canary Speech

Author: Glenn Beck

 

This is NuCalm, the show for those looking to improve sleep quality, manage stress, and boost recovery. Brought to you by Solace Lifesciences, the makers of NuCalm, the world’s only patented and proven neuroscience technology that works within minutes, without drugs, every time! In over one million medical sessions, NuCalm has helped men and women around the world.

 

NuCalm: stress relief for the way we live today, technology to help you disconnect.


Full Transcript

David Poole
This is David Poole with Solace Lifesciences. I’m here with Henry O’Connell, the CEO and President of Canary Speech, and Erica Robinson who runs our Research Department and heads up the Mastermind Project for Solace. And we’re going to speak to Henry today about the Canary Speech App. How he got involved in the business, why his tech is so powerful and potent and best in class, and then how he uses it. And then we expect to use it as the embedded technology in the NuCalm platform to analyze stress and rests. Erica is going to be leading the questions and Henry’s got all the knowledge we want.

Erica Robinson
We’re happy to have you. Thanks for starting us off, David. I think a good place to start, Henry, would be, could you just tell us a little bit about the history of Canary Speech, how it all came about, how long the company’s been around and the technology? Just to give us a little bit of an overview.

Henry O’Connell
Sure. Thank you. So the company was founded about five years ago by a very good friend of mine, Jeff Adams and I, founded the company. Jeff and I met nearly 30 years ago. I was at the National Institutes of Health as a Researcher in Neurology. Our group specifically studied neurological diseases and Jeff Adams, had a much cooler job, he was at our NSA, our spy group, where he did decoding of encrypted messages during the cold war. He’s a mathematician by training. From there, after I left NIH, we both were at NIH and the NSA for about five years.

Henry O’Connell
I left and went on into technology companies and Jeff left and went on into a career of speech and language. Jeff Adams, my business partner, is responsible for most of the major advancements in speech and language. He is so well-known that when Amazon was seeking to build the Alexa, the Echo Project, they simply bought the company Jeff was in, and Jeff then led the team at Amazon that three years later launched the Amazon Echo. Along with that was the patent lawyer, Jeff O’Neill and 17 of his speech and language scientists at that company, when they concluded that work about five years ago, Jeff and I formed Canary Speech.

Henry O’Connell
Jeff O’Neill, the patent lawyer, Jeff Adams, myself, and the core group of speech and language scientists came together to create a new generation of speech analysis for human condition and disease. We had been awarded five patents, two in Europe and three in the US, we’re pending patents in China, Japan, and Korea. Our initial patents, both in Europe and US received a hundred percent of their claims, which is unique, truly.

Henry O’Connell
Our approach was to look at the underlying bio features and markers in speech. So below the level of word, we analyze 2,548 speech features for human conditions. Things like stress and anxiety, depression, tiredness, alertness, cognitive function, and Canary has been able to build mathematical models that can very accurately analyze conversational speech, three to four minutes of speech. And we can provide a score for an individual that represents where their level of stress is.

Erica Robinson
That’s amazing. There’s definitely a need for this in society, especially, everything’s kind of moving towards that app platform and remote health care, I think. It’s really amazing. How does the technology compare to other voice recognition softwares or technologies in the healthcare field and in the market?

Henry O’Connell
Sure. Most analysis, currently in the field, are being done by word pattern. So they look at what words you’re speaking. They look at the gaps between words or filler words to try to analyze stress patterns. So do you speak stress related type of words? One of the problems with that is that it’s highly related to the particular characteristics of a person’s vocabulary and the use of words that they have, which may be affected by their education or by their culture or by their community.

Henry O’Connell
And in our case, we look at sub units below that and what that offers us quite simply as every 20 milliseconds, we look at all 2,548 bio features, and we extract those. We slide that window 10 milliseconds and do the same thing. So in a three minute audio, we have approximately 200,000 data points, correspondingly using other technique, you might have a thousand data points. And the consequences are when you’re looking at 200,000 data points, you can be more accurate, you can be more specific. We can tell the severity of someone’s stress in the morning, and after doing a meditation, or an event where they’re calming themselves, after they go walking, after they exercise, they could measure again and see the impact of that in their lives, providing them with biofeedback that’s accurate and objective.

Erica Robinson
And does this work for any language or currently kind of only for the English language? So could you detect stress and anxiety and tiredness across all languages?

Henry O’Connell
Currently we have validated our technology in Japanese and English. We are starting a project in mainland China in Mandarin Chinese, and in Europe. We are … when I say validated, we’ve actually taken it out because we’re sub-word level, we tend to be language agnostic. So it doesn’t matter what language we’re speaking. We will be adding French and Spanish next, in studies, we have a project in Dublin, Ireland, so accents and things like that really don’t affect our measurements, a Boston accent versus say an accent from South Carolina, it doesn’t matter.

Erica Robinson
That’s really, really cool. How would you train your algorithms or do the machine learning? How does that work? So I know with the Canary Speech that we’re partnering with NuCalm, we go through some of the GAD and general anxiety detection scores, and the stress level scores. Is that what you’re using to basically train the software?

Henry O’Connell
Yes, that’s accurate from the very beginning, Canary recognized that in order to accurately extract features that were specifically related to stress or anxiety, we needed to do that within controlled studies. So we’ve conducted dozens of controlled studies with pharma and with clinicals where we’ve used GAD-7, Pantis, Stay, the Modrus Test. For cognitive we’ve used things like the Montreal Protocol, which is used in assessing cognitive function and for Alzheimer’s studies. So we’ve aligned our feature extraction and our model building with industry standards that are accepted in clinical environments.

Henry O’Connell
And that’s provided us with the ability to extract specific features in established models. So to give you an idea of with stress, based on those clinical trials, we extracted 26 bio-features of the specific 2,548. So those 26 correlate very highly with stress. I like to tell people if we made it 27, the model isn’t any better and if we made a 25, the model isn’t as good. So 26 using machine learning and AI allowed us to create this algorithm. That algorithm, then when you pass three minutes of speech through, it can provide an accurate score back to the person on what their stress levels are. Now, over time, those models get better and better. They keep refining themselves and retraining themselves with exposure to a community for instance.

Erica Robinson
So the scores that you’re obtaining at the end of a voice recording, are they compared to your own baseline or to the community norm or both?

Henry O’Connell
Really both. So the score can establish a baseline for an individual. So when they see a score, when they first take it, the algorithm actually will become accustomed to them. And the score becomes related specifically to them, it also will compare against the community. So for instance, with NuCalm, there’s a community of users. We’ll establish what the normal level of stress or anxiety or tiredness or restfulness is for that community, and they can potentially compare themselves with the community that they interact with regularly. We also compare against standards and norms in the industry. So you have a standard comparison against the vast community and two dozen years of research. And then you have the individual, and also how they compare with the community that they’re engaged with.

David Poole
Henry, may I ask, sorry, Erica.

Erica Robinson
No, go ahead.

David Poole
How is this compared to some of the usual suspects in diagnostics for stress like cortisol testing, saliva testing, galvanic skin response, HRV, blood pressure, heart rate, those kind of things. Where does yours fit in there? Is it more accurate, less accurate?

Henry O’Connell
Well, our test takes into account, literally, hundreds of thousands of data points. Speech and language is considered to be the densest and most complex source of data that the human body produces. I like to say, and I’m a biochemist and a genomicist and I’d like to say next to the DNA only. So next to our own DNA, speech is probably the richest source of data we produce. We can gather a range of information from facial expressions to eye movement, to heart rate, to cortisol testing. All of those things can be additive to any other measurement that we might take.

Henry O’Connell
However, speech has some real advantages. One it’s very, very dense in data and information. And secondly, it’s easy to gather. We’re able to do an analysis on a speech pattern of about three minutes in milliseconds, and we can return a score then to an individual within two or three seconds. If you do a cortisol test, of course, you have to wait a while to get the results. If you do a hair cortisol test, it’s a little different than a saliva test. If you’re measuring your heart rate, that could be pretty instantaneous, but it can be affected by many things. It can be affected by exercise. It can be affected by the moment, there’s a fight or flight response that’s affected by the release of enzymes, which is what cortisol is measuring. All of those things have multiple reasons for being engaged in the body. Where speech itself has connected directly to the central nervous system.

Henry O’Connell
And the central nervous system then controls our muscular system, our vocal cords, our breathing, all of those things come into play along with how we form language and the wonderful thing about this, and we were talking about this earlier, Erica, is the human brain is capable of doing that, not in one language, but in hundreds and hundreds of languages and those languages communicate, not only words and meaning, but they communicate emotions. All we have to do is talk to any of our friends to realize that the words they’re saying are not just the meaning, but also the emotion they’re conveying in those words. And so language communicates both the literal meaning of what we’re saying from the words, but also the emotions that we have added to that because of who we are as people.

Erica Robinson
That’s very powerful. So you’re going on about emotions. Could you tell us, and you talked a little bit that you’ve done some research with Alzheimer’s in the mental health field, what are some of the research studies that you have going on? Or what are some of the patents, if you have any, in that kind of field? Because I know that’s a big point of discussion in today’s society.

Henry O’Connell
Yeah. Most of our studies today are on the area of mental health. So in mental health, we generally look at two areas. We look at what are called mood disorders or thought disorders and mood disorders are things like depression and stress and anxiety and thought disorders or things like bipolar and schizophrenia and things like that. What we found in a range of studies in depression, anxiety and stress was that we were able to selectively differentiate between the anxiety a person was experience and the level of their depression. Other factors that are key and important are how rested an individual is because of one has a prolonged period over which they don’t sleep well. That impacts on both their mood, depression, anxiety, and stress, but it also impacts on cognitive function. We don’t think as well. And we don’t think as clearly, and we can’t do computations and thought patterns as well, if we’re overly tired.

Henry O’Connell
So one of the wonderful things about our technology is that in that same three minutes of speech, if we were looking at depression, we can look at what is called comorbidities in the clinical area, which are associated or related human conditions or disease. So we look at depression, but at the same time, we may look at restfulness as well as stress and anxiety. An individual may be depressed, not because they’re chronically depressed, but because they’re exhausted. And so an individual looking at these measurements, which could all be assessed during the same three minutes of speech, could say to themselves, I need to get better sleep. I need to concentrate on getting my sleep improved because that’s going to impact on lowering my depression and managing my stress. Our patents today relate to, broadly speaking, they relate to the use of these biomarkers that we have identified within conversational speech, using smart devices, such as smartphones and tablets and things like that, in conversational speech to assess this.

Erica Robinson
And if we could go back to you were talking about people being able to understand the comorbidities of their mood disorder. Is that something that a consumer could learn themselves? Or is this something that a clinician would use to be able to explain to their patients?

Henry O’Connell
Oh, I think a consumer can use that to help manage their own health status and health condition, with respect to just their general wellness. If they find themselves creating a history over a period of weeks, and months of what their level of stress and anxiety is, and also they pay attention to, and we can measure their restfulness, if they find that in periods where they are simply not as well rested, they’re tired, for instance, they’re not getting good sleep or they’re overworking, or they’re traveling too much. Those things can all contribute to poor sleep and therefore a lack of restfulness, of tiredness in the person. They’ll probably see a corresponding pattern to increased stress and anxiety. So they can see these patterns and they recognize that as their stress and anxiety is going up, they have to pay attention.

Henry O’Connell
They just simply have to pay attention to getting better sleep, maybe delaying the next trip or getting to bed a bit earlier. And recognizing that part of their health and wellness is related to good sleep. As we all know, our nervous system wears out in just normal use. Most of its repair, the vast majority of its repair is done during deep sleep. And the consequences are that if we don’t have that deep sleep for prolonged periods of time, our nervous system is not repairing itself at the neuron level. And the consequences are that we’re more nervous, we’re more anxious, and we need to allow ourselves to have a healthy life and repair ourselves. So the combination of that information can help an individual manage their lives. That information, in a clinical setting, may help a doctor realize I don’t need to prescribe depression medication. I need to prescribe a pattern of life and living that allows this person to get better rest and sleep.

Erica Robinson
And have you seen that being used in both of those ways? And could you maybe give us a little bit of your personal story, because we do know that you’re using the technology. You said you and your wife are using it. Could you give us a little just anecdote about how you use it day to day and how it does impact your life?

Henry O’Connell
Sure. For the last year and a half, my wife and I daily measure our stress, anxiety and tiredness. Also, we look at cognitive function and it’s interesting because each individual is going to establish where their own baselines are and you have to be careful. It’s not a competition. So I tend to have lower stress levels than my wife does, I just tend to, about 10% lower. And that’s a normal. When we first started doing it and stress levels were higher, it was almost a competition. “Well, why are yours higher? You don’t get as much sleep as I do.” And what we found was that as we establish what our personal baselines were, we looked more within ourselves. How am I comparing to where I was last week? And how does that compare to the fact that I’m not getting as much rest and sleep as I know I need?

Henry O’Connell
And for every individual, the amount of sleep they need and the amount of rest they need is a little different too. But those things will correlate with their level of stress and anxiety. So they can begin to get those things in balance for themselves and for their own lives. And then my wife and I compare, not every day, but we compare our scores with one another. And I have to be careful because if I’ve been a little bit testy, my wife will say, “I bet your scores are higher.” And she’s usually right.

Henry O’Connell
And so it gives us feedback in our own lives. That allows us to say, “I don’t need to watch this episode of whatever on TV. I probably could just go to bed now.” I find in my personal life, I exercise quite a bit. I like to get out for a walk or a run every day. And if life is very busy, and our lives are, I tend to let those things slide. But if you’re measuring them regularly and you’re seeing an impact in your life, that’s not what you want. Then it draws your attention to the reality that, all of these things are quite important.

Erica Robinson
That’s amazing. David, do you have any other questions?

David Poole
Oh, I could talk all day. I like hearing Henry speak. You’ve got a very soothing voice. My stress level’s going down, just listening to you, Henry.

Henry O’Connell
It’s the Irish in me, David.

David Poole
I get it.

Henry O’Connell
I think I told you, I took the Ancestry DNA and I got a letter from them. They actually wrote me a letter, sent me a letter and I go to the mailbox it’s Ancestry, and I thought it was an advertisement to buy another thing or something. I opened it up and they said, “Mr. O’Connell, we wanted to recognize that having reviewed your DNA test, we found at 99.2%, you were more Irish than our employees in our Dublin office.” Which I thought … my four grandparents are immigrants from Ireland. But I thought there’s got to be something else in me, but apparently not.

Erica Robinson
Apparently not. I haven’t done one of those yet. Maybe at some point.

Henry O’Connell
It’s fun.

Erica Robinson
So just before we got on the recording, you were telling us about some of the new research projects you’re working on with Canary Speech and some of the new opportunities that you have within the United States. Do you think you could share those with us, again?

Henry O’Connell
I would love too. Canary Speech has been able to associate with some wonderful organizations, clinical centers across the country and join in a partnership with them to bring speech and language assessment, to help aid in evaluating their patients, doing assessments on stress and anxiety. Think of it as almost a mental health assessment. So a person comes in and they don’t necessarily think their life is any different than anyone else’s. But in fact, their stress levels are 40% higher than the normal, but they would report to their doctor, “Life is life, I’m okay.” But if they could begin to manage these things, either in their own personal lives or get counsel from their clinician about changes in their behaviors to help manage stress, it could make a significant improvement.

Henry O’Connell
Stress is a terrible thing. Over time, it impacts on the nervous system, the immune system, it destroys our immune system, if prolonged stress is allowed to be pervasive in our lives. So some of the practice we’re doing, we are working on a wonderful project and Alzheimer’s with the Gates Foundation and the Alzheimer’s Association. That’s going to be a multi-site, multi-country evaluation using speech and language as a tool to both assess the effectiveness of therapeutics that are being used, as well as individuals to personally manage their lives and clinicians to diagnose severities and progression of the disease. So that’s an amazingly wonderful thing. We’re working with the state of Wyoming for remote health and stress and anxiety in older populations. How are they progressing cognitively as they age? Doing assessments that are simple and easy to do and can be done at home. Clinicians can then receive those assessments and help them manage their lives better and improve, we hope, the quality of care.

Henry O’Connell
We’re engaged with a number of hospitals. As I mentioned, just wonderful ones, Intermountain Hospital, Hackensack Meridian, Atrium Hospital, and others that are just professionals, wonderful professionals. And there’s just a lot of desire, I think, to improve the effectiveness of the clinical teams we have, to provide care to remote populations as well as to better serve the communities that are around their hospitals. And there’s a real desire to help people get the information they need to manage their own lives too. And because most of our lives, something like 0.06% of our lives, is ever spent with a clinical person. So the vast majority of our lives, 99% plus of it, is something that we need to be able to have information, be able to manage our own lives. So that’s, I think, Canary can contribute to that. We were talking about earlier about heart rate and cortisol tests or other things. Speech is another element, and I think a wonderful one, but it’s another element that can provide information to individuals, to self manage their lives.

Erica Robinson
That’s a really good end note right there. O’Connell tells us about the future of Canary Speech and the potential for this technology, all throughout the world really.

Henry O’Connell
Well, I appreciate it. I appreciate the time today, very much.

Erica Robinson
I really do too. And just one little fun thing that Dave and I were talking about, just to get to know you a little bit better, we wanted to just ask a personal question. What was the last book you read?

Henry O’Connell
Well, I really … and my family tells me I should stay away from it, but I really enjoy reading about politics right now. And I was reading a Glenn Beck book, it’s not a new one of his, just the other day. And honestly, I just opened it up and started reading. I don’t remember what the title is, but it was Glenn Beck. And he has softened, in the last couple of years. Five years ago, my stress levels would go up reading him, but he has softened in the last couple of years, particularly. And he was talking about the need for all of us to cross aisles and work together. And this particular book was on some of the political framework of today, which is complex in the US, as you know.

Erica Robinson
That’s awesome. Okay.

Henry O’Connell
Thank you.

Erica Robinson
Thank you so much for taking some time with us. We’re really excited to get all of our Mastermind participants using the technology and providing data points to really build the machine learning software. And we’re excited to have you on a call with everyone, hopefully answering some more questions that come up from our consumers themselves.

Henry O’Connell
We look forward to it, Erica. Thank you.