Smarter Measurements for Smart Speakers
In this article, we describe techniques to characterize the frequency response, output level, and distortion of the device under test to enable direct comparisons between Internet of Things (IoT) smart speakers and conventional speakers.
Author: Daniel Knighten (Listen, Inc) and Glenn Hess (Indy Acoustic Research). Reprinted from the March 2018 issue of Audio Xpress.
Smart Speaker Audio Performance Testing
Sources:
- Knighten, Daniel, and Glenn Hess. “Smarter Measurements for Smart Speakers.” Practical Test & Measurement, Mar. 2018.
Main Themes:
- Rapid Adoption & Market Growth: Smart speaker adoption is skyrocketing, with predictions exceeding 100 million units shipped by 2024. Audio quality is a key differentiator in this competitive market.
- Unique Testing Challenges: Evaluating smart speaker audio performance presents unique challenges compared to traditional speakers due to the lack of direct input/output paths, voice activation, and reliance on cloud-based services.
- Necessity of Objective Measurements: Objective measurements of frequency response, output level, and distortion are crucial for comparing IoT smart speakers to conventional speakers and ensuring consistent quality.
Important Ideas & Facts:
- Cloud-Based Operation: Smart speakers lack direct input/output paths, relying on internet services for audio playback and recording. This necessitates retrieving recorded signals from online platforms and uploading test signals to cloud music services.
- Voice Activation & Control: Testing requires overcoming the challenge of voice activation. This involves understanding each device’s activation phrase and recording time to ensure accurate signal capture.
- Sampling Rate Error: Asynchronous audio operation in smart speakers can lead to sampling rate discrepancies between the device and the measurement system. Algorithms are required to correct these errors and ensure accurate analysis.
- Test Signal Design & Analysis: Specialized test signals, such as stepped sine sweeps, are employed to evaluate frequency response and distortion. Analysis utilizes heterodyne filtering to extract fundamental and harmonic distortion products.
- Measurement Setup: The authors recommend adapting existing standards like IEEE 1329-2010 for speakerphone testing due to similarities in use cases. Measurements are ideally conducted in an anechoic chamber to minimize environmental interference.
- Future Research Areas: Further research is needed to develop tests involving dynamic signals like noise, modulated noise, and real-world speech and music. Evaluating activation phrase detection amidst noise and multiple talkers is also crucial.
Conclusion:
Developing robust and standardized testing methodologies for smart speaker audio performance is crucial for ensuring quality and driving innovation in this rapidly evolving market. Overcoming the unique challenges posed by cloud-based operation, voice activation, and asynchronous audio necessitates specialized test signals, analysis techniques, and measurement setups. Further research is needed to encompass real-world scenarios and evaluate more complex aspects of smart speaker performance.




