Is_my_Sleep_sound_data_collected_transparently_and_securely_Mel_Spectrogram
Sleep & Tech

Is my sleep sound data collected transparently and securely?

|by Asleep

    [Editor’s Note] 

    Asleep's Sleep AI analyses users' sleep through sleep sounds. Sleep sounds are various noises that occur during sleep, such as breathing and tossing and turning. Asleep's Sleep AI collects sleep sounds through the microphone of a smartphone or smart device entrusted to it, monitors the user's sleep, and provides real-time measurement results.

    Because sleep sounds are personal information, they need to be managed carefully. In particular, although wearable sleep trackers are overwhelmingly convenient, the process of collecting and analysing personal information such as sleep sounds through the smartphone microphone during 7-8 hours of sleep every night cannot ignore the aspect of 'security'.

    How does Asleep collect and analyse sleep sound data? Is my sleep sound data managed securely? Explore Asleep's data security policies and principles, which are considered crucial alongside convenience and accuracy.

    Asleep's Sleep AI uses the microphone attached to a smartphone or smart device, without the need for additional devices, to analyse sleep through various sounds that occur during sleep, such as breathing and snoring. It tracks sleep stages, detects the stability of breathing and easily identifies sleep patterns such as snoring.

    Breathing is a crucial factor in understanding sleep. In particular, sleep stages, which are essential information for assessing sleep health, vary according to the activation levels of the central nervous system and the motor nervous system. As breathing changes in tandem, analysing breathing effectively allows the identification of sleep stages.

    The state of the central nervous system can be assessed through the rhythm, cycle and pattern of breathing. When assessing the state of the motor nervous system, breathing is also an important indicator. This is because tracking the tone and depth of breathing allows the state of the motor nervous system to be monitored. In addition, by understanding how breathing sounds and patterns change, it is possible to detect conditions such as sleep apnea or hypopnea during sleep.

    💡 Want to know why accurately measuring and analysing your sleep stages is so important to understanding your sleep health?
    👉🏻 
    <Sleep stages: what are they and why are they important?>

    Asleep's AI-based sleep analysis technology has been developed using state-of-the-art machine learning algorithms trained on extensive sleep data. The breathing sounds used in this technology are measured using the microphone built into smartphones. As a result, people can track their sleep at home for several days using just a smartphone, without having to go to a hospital. In addition, all sleep tracking information is recorded in an easy-to-understand format, allowing for regular tracking and management of sleep.

    During the process of collecting and analysing sleep sound data, Asleep does not collect or analyse raw sound data. The sleep sounds captured by the smartphone microphone are converted into an image format called a Mel Spectrogram using a code developed by Asleep. This Mel Spectrogram is then sent to Asleep's sleep analysis server. The Mel Spectrogram, as shown in the image below, visualises the waveform of the sound in frequency intervals known as the Mel Scale. In other words, sleep sound data containing personal information is not stored on Asleep's servers or the servers of services using Asleep's sleep AI technology.

    In addition, Asleep removes sensitive sound information and personal data generated in the sleep environment and processes it in a way that makes reverse transformation impossible. When sleep sounds are transformed into Mel Spectrogram images, the resolution is deliberately kept low enough that specific conversations or the identity of the person speaking cannot be discerned. However, the image resolution is still sufficient to analyse important information for sleep analysis, such as the tone and pattern of breathing. This ensures the accuracy of the AI analysis while protecting personal information.

    As part of the process, the resulting Mel Spectrogram is sent to the Asleep sleep analysis server every 5 minutes. The sleep sounds captured by the smartphone's microphone are transformed into Mel Spectrogram images every 30 seconds, and these transformed images are sent to the sleep analysis server every 5 minutes.

    Asleep chose a minimum 30-second unit for the Mel Spectrogram because it is consistent with the analysis time for sleep stage information provided by polysomnography(PSG) and the minimum time unit defined by the American Academy of Sleep Medicine(AASM) for analysis of sleep stage information. Asleep performs sleep stage analysis according to the most accurate medical sleep study and internationally recognised standards, ensuring the highest level of accuracy.

    🌞 Click on the image below to see a Mel Spectrogram image of the sounds of each stage of sleep and to hear the subtle changes in breathing that occur during each stage of sleep.

    In this way, Asleep remains free from privacy concerns by retaining and using only the information necessary to measure sleep. It accurately analyses an individual's sleep sound data, while at the same time collecting data in a safer way using the Mel Spectrogram transformation method.

    🔐 Asleep keep your personal information safe.

    1️⃣ The sleep sound data collected by the smartphone's microphone is transformed into an image format called a Mel Spectrogram and then transmitted to the Asleep AI analysis server.
    2️⃣ The transformation is done at a resolution sufficient to analyse sleep stages and phenomena that occur during sleep, making it impossible to recover the original sound.
    3️⃣ Sleep sounds containing the user's personal data are not stored on Asleep's servers or on the servers of services using Asleep's sleep AI technology.
    📍 Want to know more about how our Sleep AI technology detects differences in breathing during different stages of sleep?
    👉🏻 <The most noted Sleep AI Technology in World Sleep Congress 2023>