Overview
ISO/IEC 30122-2:2017 - "Information technology - User interfaces - Voice commands - Part 2: Constructing and testing" specifies technical criteria and test methods for voice commands and the speech recognition engines that recognize them. The standard focuses on the phonetic design of spoken words and phrases used as voice commands and on standardized procedures to verify recognition performance under realistic conditions. Conformity requires meeting all requirements in Clause 5 of the standard.
Key topics and technical requirements
- Phonetic requirements
- Voice command phrases should favor plural words and consist of plural syllables to improve recognizability.
- Pronunciations must be documented using a phonetic markup (e.g., IPA, SAMPA) to reduce ambiguity.
- All significant pronunciation variations (more than two pronunciations) must be registered to the voice command attribute.
- Design principles
- Commands should be easy to pronounce by native speakers and usable by people with speech difficulties.
- Phrase length should be short enough to be spoken in one breath; avoid tongue‑twister combinations.
- Testing speech recognition engines
- Subjects: minimum 10 participants (ideally ≥100) drawn from diverse demographics (age, gender, native language, culture, disabilities, geographic location). Include users with varying experience levels and blind/visually impaired groups as appropriate.
- Samples: at least 10 command samples including frequent, critical, difficult-to-recognize, and random commands.
- Conditions: tests must reflect realistic usage (e.g., in‑car noise levels for navigation systems).
- Procedure: multiple trials per sample per subject; samples may be live or recorded for later processing.
- Metrics: report test conditions and use standard indexes:
- Sentence recognition ratio = (correctly recognized sentences / total tested sentences) × 100%
- Word recognition ratio = (correctly recognized words / total tested words) × 100%
- Word accuracy = ((correctly recognized words − inserted error words) / total tested words) × 100%
Applications
- Design and evaluation of voice user interfaces (VUI) for consumer electronics, automotive systems, smart home devices, and ICT applications.
- Development and QA of speech recognition engines and voice command grammars.
- Accessibility testing for assistive technologies used by people with speech or mobility impairments.
- Procurement and compliance checks where standardized voice command behavior is required.
Who should use this standard
- UX and VUI designers, speech engineers, QA/test laboratories, product managers for voice-enabled products, accessibility specialists, and system integrators.
Related standards
- ISO/IEC 30122-1 - Framework and general guidance for voice commands
- Referenced ergonomics and language resource standards in the ISO 9241 and ISO 246xx series for complementary guidance.
Keywords: ISO/IEC 30122-2:2017, voice commands, speech recognition engine, voice command testing, phonetic requirements, voice user interface (VUI), speech recognition testing.