Table 5.

Summary of task-specific features for automated quality assessment.

Task	Feature categories	Rationale
Speech	Speaker composition (multispeaker detection) Linguistic adherence (similarity, unique words) Temporal structure (audio and silence durations) Acoustic clarity (signal-to-noise ratio)	Capture deviations from the target sentence, missing or noisy speech, and multispeaker recordings that compromise intelligibility.
Smile	Face visibility (faces per frame, detection rate) Camera framing (waist/knee visibility) Lighting and reflection (eye detection, eyeglass ratio, brightness metrics) Task timing (smile onset/offset, idle time, smile count) Device context (aspect ratio)	Identify recordings with missing or multiple faces, poor illumination or eyeglass reflections, improper distance from the camera, and deviations from the 3-smile protocol.
Finger-tapping	Hand visibility (multihand ratio, wrist missing ratio, idle time) Movement compliance (tap count, duration, speed) Lighting (frame/hand brightness ratio) Camera framing (waist/knee visibility) Recording metadata (video duration, aspect ratio)	Detect recordings with missing or multiple hands, incorrect tap counts or speed, poor lighting, excessive distance from the camera, or nonstandard recording setups.