EMBEDDED VOICE-COMMAND TRIAL RECORD
Yenra - October 4, 2026
Guide: https://yenra.com/speech-recognition/

Purpose: evaluate a complete voice-control task and its recovery path.
Edit in a text editor; repeat the trial fields for each attempt.
Use a low-consequence task in an authorized test setup. Agree on consent,
privacy, breaks and recording with participants. Stop when requested.

CONFIGURATION
Device, microphone and enclosure:
Firmware, recognizer/model and language:
Settings and required training:
Intended task and alternative input:
Distance/orientation, background sound and device audio:
Network state; initial setup or restart conditions:
Acceptance criteria agreed before testing:

ONE TRIAL
Trial identifier and date:
Participant identifier (avoid unnecessary personal information):
Intended command or explicitly designated non-command condition:
Expected result:
Wake detection observed:
Recognized command, if observable:
Actual action and feedback:
Classification: correct / missed / wrong action / false activation
Correction, help or alternative input needed:
Completed after recovery? User's effort/preference:

INVENTED SUMMARY AND ARITHMETIC
Intended-command trials: 10
Correct first-attempt actions: 8
Missed requests: 1
Wrong actions: 1
Check: 8 + 1 + 1 = 10; correct first-attempt result = 8/10 = 80%.
Separate non-command trials: 10
False activations within those trials: 1
This is one false activation in ten defined non-command trials.
It is NOT an hourly false-activation rate. Continuous observation needs
recorded duration and a defined event-counting method.
These invented counts are not a product result or an acceptance threshold.

DECISION
Results by participant and condition (keep alongside totals):
Stage to investigate:
One proposed change and owner:
Retest setup and date:

Source context: the guide links to Sensory SDK documentation and W3C's
draft voice-systems research. This worksheet is an editorial test aid,
not a standardized benchmark or accessibility-conformance certificate.