Unobtrusive Sleep Position Classification Using a Novel Optical Tactile Sensor.

Area of Science:

• Biomedical Engineering
• Sensor Technology
• Sleep Science

Background:

• Accurate sleep position classification is vital for sleep monitoring and interventions.
• Current methods may be obtrusive or lack precision.
• Positional Obstructive Sleep Apnea is a prevalent condition often undertreated due to CPAP intolerance.

Purpose of the Study:

• To introduce a novel, unobtrusive under-mattress optical tactile sensor for classifying sleep positions.
• To evaluate the sensor's sensitivity and accuracy in detecting pressure and shear forces.
• To demonstrate the feasibility of using this sensor for robotic bed applications in sleep apnea treatment.

Main Methods:

• Developed an under-mattress optical tactile sensor using a camera to track particles in a silicone layer.
• Characterized sensor sensitivity by applying various weights (258g-5000g) under a mattress.
• Collected sleep position data (supine, lateral left/right, prone) from a human subject.
• Trained a convolutional neural network with residual blocks for sleep position classification.

Main Results:

• The optical tactile sensor demonstrated high sensitivity, detecting a 258g weight even under a mattress.
• A neural network achieved 91.2% accuracy in classifying four distinct sleep positions.
• The sensor reliably inferred pressure and shear distributions from silicone deformation.

Conclusions:

• The novel optical tactile sensor is highly sensitive and accurate for unobtrusive sleep position classification.
• This technology shows significant promise for integration into robotic beds for sleep disorder management.
• The sensor's scalability and concealed nature make it a viable solution for advanced sleep monitoring and intervention systems.