04.03 + 04.05: Narrowing In
This week we dove further into primary and secondary research and decided to narrow our scope. Based on the feedback from our last progress presentation combined with our own research, we found that engaging with stakeholders in the hospitality industry could prove more fruitful than attempting to take on the challenges inherent in a more medical/clinical service. We also discovered opportunities to expand our application of music as a tool for wellness to include not only sleep, but the evening and morning routines surrounding a good night’s rest.
Service Statement
Rhythms is a service that offers improved sleep quality at home and while traveling by intelligently regulating your sleep regimen through procedurally generated music. Rhythms is unlike Brain.fm, Headspace, and FitBit, which only provide individual components of this service.
Service Blueprinting
So far, we have nailed down our top level user actions for for our primary user, who is registered with our service, and for our guest user, who interacts with our service for the first time in a hotel.
Travel, Stress + Sleep Research
Can a travel app make you mindful and relaxed?
This CNN article addresses some of the psychological and physiological effects travel can have on an individual and briefly reviews a series of meditation platforms; some are specifically tailored for travelers and others are more multipurpose.
Headspace also addresses the impact of travel on a person’s sleep quality and overall wellness in its article The Sleep Doctor on How to Beat Jet-Lag. Jet-lag — and more generally any travel that involves a disruption in normal circadian rhythms — can be highly detrimental to mental and physical health.
(More) Competitive Analysis
(More) AI Generated Music Research
The use of AI generated music in our service creates some interesting opportunities and advantages. First, it removes the need for us to worry about licensing music for our service. If music is generated by AI software that we own, there is no need to license that music. Though, it does potentially incur some other costs that we must consider, which I will touch on later. AI generated music also allows our service to be reactive to people’s sleep patterns. Music can change and fade out as they fall asleep, come back on if they wake up in the middle of the night, and gently wake them up in the morning at the correct point in their sleep cycle. These are all things that precomposed music does not afford. The work of AI generated music in games is a perfect case study of music reacting to a situation in real-time.
When considering the technology of AI generated music for our service, there are essentially two paths to take:
“True” AI Music Composition
True AI generated music is music that is composed completely by an AI system. A good example of this is a deep-learning based system developed by AIVA. The system “learns” from compositions of famous composers and then outputs its own sheet music, which musicians can play. Feasibly, one could also “play” the music through a computer using MIDI instruments.
Though this technology is compelling, it may be more heavy handed than what we need for this project. The time needed to compose this music is likely not quick enough for a real time application. And given the fact that music that facilitates sleep is more about slow moving, consistent rhythms, dynamic compositions are not necessary. This could be accomplished with a lighter touch.
Procedurally Generated Music
A procedural approach to AI generated music can be seen in Daniel Brown’s Doctoral dissertation. He created a procedural music engine that procedurally composes music in real time by reacting to different variables in the game. He uses “theories of musical expression and semiotics in Classical and Romantic music put forth by several theorists and musicologists” to map different musical motifs to different occurrences in the game and generate changes in the music as things occur in gameplay. He uses prerecorded bits of music that get merged together in this process. For our service, this approach seems to make sense and be logistically feasible. If we have individual components of prerecorded music, we can create a system that is reactive to a persons sleep cycle and continue for as long as necessary, fade out when they’re asleep, change mood or tone if they’re tossing and turning, and more. From a business perspective, this approach would most likely mean hiring someone to compose some initial ambient music for our system to use, which will have some upfront cost. But this composition effort would likely not happen often and still be cheaper than licensing a variety of music that wouldn’t allow for procedural generation.
Examples of AI Music Services
There are some example of AI generated music in services today. Here are some examples:
Brain.fm — Using AI generated music as a tool for therapy and mental performance. It’s not clear exactly how their AI works, and they are currently trying to scientifically validate their work.
Amper — A tool that allows content creators to generate license-free music based on a variety of different settings they can control. Users can select type of mood, how long the composition needs to be, and more. Also not clear exactly how their AI works.
Jukedeck — Similar to Amper in brand positioning. Their research team updates a blog with some useful information about their work.
AI Music — Still very early stages, but AI Music is trying to create AI technologies that aid musicians in the music creation process. There is not much info available, but their website talks about dynamic co-creation of music and making music production accessible to everyone.
(More) Music + Sleep Research
Parents usually sing lullabiies to their babies to try and help them sleep. Our service is based off the proven technique that music or certain kinds of noise can help people fall asleep. Music can be used to help people with sleep disorders or sleep disruptions by increasing their sleep quality and quanity.
Through our research, it seems that there is a handful of reasons why music can help your sleep. It might be the way the music relaxes people and helps to soothe them into sleep. Music has physical effects like lowering your heart rate and slowing your breathing, both of which contribute to falling into a deeper sleep. Further, music has a direct effect on the parasympathetic nervous system. This is the part of the brain and body that helps you prepare to sleep. It has been shown that older adults, who listen to relaxing music for about 45 minutes before they go to sleep, will sleep longer, fall asleep faster, wake up less and feel that their night was more restful. Similarly, young adults expressed an improved night’s sleep when they were given the option of listening to classical music, listening to a book on tape or doing nothing and they selected listening to the classical music.
However, the results are not instant. Once you incorporate music into your bedtime routine, you need to stick to it. Positive sleep effects occur over time and it can take up to three weeks to feel the results of integrating sleep into your bedtime schedule.
Article about “How Listening to Music Could Help You Beat Insomnia”
Research has shown that listening to “self-selected” music — music of your choice — can actually shorten stage two sleep cycles. This means people reach restful REM sleep — the restorative part of our sleep — more quickly.
In the study, students who listened to 45 minutes of music before bedtime for three weeks saw a cumulative positive effect on multiple measures of sleep efficiency with similar effects reported in older citizens in Singapore. Following all this evidence, the NHS now recommends “listening to soft music” before bedtime as a method to prevent insomnia.
The first phase of our music sleep survey has been completed by 651 people, who have told us a great deal about the music that helps them to sleep. We discovered the top rated composer of sleep music in our sample is Johann Sebastian Bach. He was followed by Ed Sheeran, Wolfgang Amadeus Mozart, Brian Eno, and Coldplay.
Aside from those few top rated artists, there was an enormous variety of individual choices — with 14 different genres and 545 different artists named. And it is this data that will give us the basis to examine the features of effective sleep music. Using computer programs we will be able to pin down the consistent musical features that support sleep among these many diverse musical sounds.
The article above was reference in one of our previous Medium posts. It discusses the impacts of sleep and a study that was conducted with ninety-four 19–28 year olds. The hypothesis of the study: “Sleep disorders may result in fatigue, tiredness, depression and problems in daytime functioning. Music can reduce sympathetic nervous system activity, decrease anxiety, blood pressure, heart and respiratory rate and may have positive effects on sleep via muscle relaxation and distraction from thoughts. Control groups have not been used in most previous studies.” It resulted in the conclusion that music can have a positive impact on people’s sleep and could be an effective intervention for treating insomnia.
Expert Interview
On Thursday we had a conversation with Kristine Wilckens from the Sleep and Chronobiology Lab at Pitt/UPMC.
Kristine is a cognitive psychologist, which means her focus for sleep is more about what happens in the brain and how sleep helps to consolidate memories for the next day. We began the conversation with her by asking about her background. Her work looks at improving sleep in older adults with a focus on the quality of sleep and falling asleep. She uses techniques that improve people’s sleep depth and how continuous their sleep is. Kristine pointed out to us that longer sleep does not actually mean better sleep. Regarding her work, she is working to improve sleep efficiency, specifically for older adults, to help with memory. The goal is to have sleep with slow waves and have continuous sleep. Further, Kristine believes that music could be used to target sleep efficiency, which is how much time you spend in bed actually sleeping.
Q: How do you go about tracking people’s sleep, considering the context of a sleep lab versus being at home?
A: In a sleep lab, we use polysomnography. This method uses electrodes attached to the scale and measures brain activity (EEG), heart rate, and muscle tone. This allows researchers to identify the type of sleep the person is in.
Specific sleep stages are difficult to detect with current in-home monitoring technology, but studies often utilize an Actiwatch to track things like wake and sleep periods to determine sleep efficiency. The Actiwatches don’t measure heart rate, but can measure light. (There was also a Sensewear product on the market that measured galvanic skin responses.)
Q: Are you familiar with the potential influence of music on sleep? (We cited an article from the Journal of Advanced Nursing about college students age 19–26.)
A: I think this depends on what type of music is being played (e.g. the tempo and tones.) There are several studies now where people are using tones to entrain sleep waves. They will use acoustic stimuli with low frequency to magnify slow waves in sleep. This has been proved to improve memory. However, these studies are very specific. The researchers conducting this studies will timelock the tone to someone’s slow waves.
However, there are other studies that show using any sensory stimulus at a low frequency (1 hertz or below) makes people fall asleep or increases slow wave sleep. This could be music, puffs of air into the nose, rocking, or any type of sensory stimulation. This entrains the brain to a synchronous pattern.
Q: We’re thinking about using procedurally-generated music, fading in and out based on feedback from the sensors. Does that sound appropriate?
A: I could envision something where you have an activity monitor, connecting it in real time to the music player. If the sensor picked up only a little bit of movement, you wouldn’t want to play music. But if you have greater than, say, 2 minutes of movement past a certain threshold, then the music could play. You’d need to establish a baseline for that person to determine their normal habits.
Q: Our initial focus in designing this service was for the customer value, but we also want to consider the value of the data that would be generated from this service. Could you see that data being useful for sleep researchers?
A: Definitely, if there’s evidence that it could improve someone’s sleep, that would definitely be of interest to a lot of people. Recently there’s inceasing interest in lifestyle factors and their impact on people’s health, including sleep. Any time people can improve their sleep and health without medication or visiting a doctor, that’s valuable information.
Q: What are some common sleep disturbances that people experience? What are those factors that people could change to improve sleep quality?
A: Sleep hygiene is big consisderation. It’s important to gauge what people are doing before they go to bed — for example, people with insomnia are discouraged from reading in bed because it trains your brain to stay awake while in bed. Ther are also issues with diet (avoiding caffeine after 3pm) and environmental factors like noise from a TV or bright light from screens. These set up unproductive patterns of behavior (dealing with the suprachiasmatic nucleus.)
To wrap up our conversation with Kristine, she told us her thoughts about our service concept. Her feedback was generally very positive and she added that the most obvious consumer-based format would be an app on your phone and you put it on your bed, but sleep scientists are skeptical of the efficacy of that data collection method (specialized sensors/hardware are much more effective in gathering useful data.)
User Survey
This week we also created a user survey to gain insight into our users. Given that we’ve shifted to a direction more focused on sleep and travel, we created a survey to learn about people’s sleep quality, how that changes when they travel, and what sleep tools they are currently using.
