Summary of S4-260244: Status of the Target Device Databases

Introduction

This contribution reviews the current status of target device databases for the DaCAS-1 (Device-Agnostic Capture Audio Scene) work item. The document identifies gaps in the existing recording databases and proposes minimum requirements and timelines for completing the database collection to enable successful example solution development and evaluation.

Current Database Status

Recording Coverage Analysis

The document presents a comprehensive status table showing the availability of recordings across five target devices (A_device_A, A_device_B, B, C, D, E) and ten recording scenarios (RS 1-10):

Target Devices:
- Device A (two variants): Four-Microphone Prototype devices
- Device B: Three-Microphone Smartphone device
- Device C: Four-Microphone Smartphone device 1
- Device D: Four-Microphone Smartphone device 2
- Device E: Three-Microphone XR HMD device

Recording Scenarios (as defined in [1]):
1. SNR recordings
2. Rotational sweeps
3. Evaluation recordings based on TS 26.260
4. Fixed single talker
5. Moving single talker
6. Fixed multiple talkers
7. Fixed talker(s) and fixed music source
8. Moving single talker and moving music source
9. ISM frequency response 1
10. ISM frequency response 2

Key Findings

• Most devices only have recordings for scenarios 1 and 2 available
• Device A variants have the most comprehensive coverage (RS 1-8)
• Devices B, C, and E only have RS 1-2 recordings
• Device D has RS 1-2 and RS 4 recordings

Identified Issues and Concerns

The contribution highlights several critical limitations of the current database status:

1. Insufficient for Example Solution Development: The limited recordings (primarily RS 1-2) are adequate for device characterization but insufficient for developing, fine-tuning, and evaluating potential example solutions

2. Lack of Real-World Scenarios: Absence of recordings in real immersive usage scenarios with real (moving) talkers hinders example solution development and may result in varying quality across different target devices

3. Limited Evaluation Capability: Current database limitations prevent:

4. Subjective evaluation (even informal)
5. Successful self-evaluation
6. Optional cross-evaluation

Proposed Minimum Recording Set

Required Recordings

The document proposes the following minimum set for agreement:

1. Recording scenarios 1, 2, and 3 as defined in [1]
2. At least 2 recordings of real immersive usage scenarios (e.g., based on recording scenarios 5-8)

Recommended Recordings

For more comprehensive coverage:

1. At least 4 recordings of real immersive usage scenarios (based on RS 5-8)
2. Preferably recordings of all recording scenarios defined in [1]

Proposed Timeline

• Deadline: SA4-136 meeting for completion of minimum recording set collection
• Near-term action: For Four-microphone Prototype devices, the source (Nokia) is preparing to share RS 3 recordings at upcoming Ad-hoc telcos

Conclusion

The contribution emphasizes the urgent need to extend target device databases beyond the currently available RS 1-2 recordings. The proposed minimum recording set and timeline aim to ensure sufficient data availability for successful example solution development, fine-tuning, and evaluation across all target devices in the DaCAS-1 work item.