Signal to Noise Ratio

Definition

The Signal to Noise Ratio (SNR) is a quality metric that is calculated for every measurement. The SNR value is the relative strength of the extracted facial blood-flow ("signal") when assessed against the background ("noise"). It depends on many factors including sensor quality, lighting and motion conditions. SNR for every measurement must exceed a validated threshold before the data is processed further by the DeepAffex Cloud.

Background

SNR is affected by many factors, some of which are listed below in the order of significance:

• Camera sensor quality
• Distance from camera (i.e., pixels per mm of skin)
• Movement
• Lighting
• Skin tone
• Obstructions on the face such as makeup, sunscreen, eye-glasses, etc.
• Camera lens quality and cleanliness

Interpretation

SNR values less than 0.5dB are considered to be too low for reliable estimation of other signals. In these circumstances, the DeepAffex Cloud doesn't compute any other signals present in the Study and returns a sentinel value of -100dB.

For most applications that perform 30-second measurements using 5-second chunks, we recommend the following rules to cancel a measurement early based on SNR:

• If SNR < 0.5 at 15 seconds (after three 5-second chunks) OR
• If SNR < 0.5 for two of the remaining three 5-second chunks
  • Cancel measurement
• If SNR < 0.5 for the last chunk
  • Do not display measurement results