Report from Audio SWG Teleconference on DaCAS (13 January 2026)

Meeting Overview

The Audio SWG teleconference on DaCAS (Diverse audio CApturing system for Smartphone devices) was held on 13 January 2026 with 20 participants for 1 hour. The meeting processed two input documents and produced two agreed output documents with revisions to be implemented in DaCAS-2 and DaCAS-3.

DaCAS Example Solution Evaluation Procedures (S4aA260001 → S4aA260008)

Contributors: Nokia, Fraunhofer IIS, Dolby Laboratories

Main Technical Contributions

This contribution established the framework for evaluating DaCAS example solutions through a combined approach:

Evaluation Methodology

• Combined evaluation approach comprising both self-evaluation and cross-evaluation
• Cross-evaluation as optional task depending on available effort from cross-evaluation laboratories
• Three-tiered cross-evaluation approach (hierarchical structure):
• Minimum requirement: Cross-evaluation based on output signals provided by solution proponents
• Cross-evaluation using input signals and upmix methods
• Running executables to generate outputs for evaluation

Key Discussion Points and Resolutions

Cross-Evaluation Flexibility:
- Cross-evaluation labs have freedom to conduct evaluation based on already provided signals or process evaluation signals with submitted solutions
- The approach depends on available effort and resources at the cross-evaluation laboratory

Dataset Requirements:
- Mandatory minimum evaluation dataset to be defined
- Cross-evaluation limited to datasets that solution proponents have provided
- Proponents must process recordings from the evaluation database
- Cross-evaluation labs may use mandatory dataset and potentially additional signals

Clarifications Added:
- Editor's note added to Clause 2.1 regarding cross-evaluation procedures
- Clarification that recordings shared in the database should be used for processing with example solutions
- Any company (proponent) can conduct cross-evaluation if desired

Agreement Status

The revised document (S4aA260008) was agreed with changes in brackets to be implemented in DaCAS-2.

DaCAS Example Solution Deliverables (S4aA260002 → S4aA260009)

Contributors: Bytedance, Beijing Xiaomi Mobile Software

Main Technical Contributions

This contribution defined the package and deliverables required for DaCAS example solutions.

Deliverable Components

1. Target Device Information:
- List of supported target devices (revised from "target device to provide")
- Device specifications and characteristics

2. Dataset Description:
- Clarification that this refers to recordings used to develop example solutions
- Discussion on whether to include external data used for development
- Distinction between:
- Common database for algorithmic development (transparent, shared data)
- External data (e.g., for neural network training)
- Geometric/analytical data (part of algorithmic description)

3. Algorithmic Description:
- High-level description of solution development approach
- Differentiation between signal processing and neural network solutions
- For signal processing: description of methods (e.g., FEM method, upmix matrices)
- For neural networks: indication of training data usage

Key Discussion Points and Resolutions

External Data Usage:
- Proponents allowed to use external data (e.g., for neural network development)
- Current common database amount may be insufficient for neural network training
- No prohibition on external data, but transparency required
- Common database serves as baseline, not as closed/exclusive dataset

Unified Deliverables:
- Suggestion to combine deliverables for signal processing and neural network solutions
- Recognition that solutions may combine both approaches
- Need for unified set of deliverables across solution types

Documentation Requirements:
- Code package contents to be specified
- Similar to table format used for dataset descriptions
- High-level description of how solution was developed
- Reference to potential datasets used

Outstanding Items:
- Bullet 2 (dataset description) placed in brackets for further discussion
- Need to combine and unify deliverable descriptions
- Editor notes inserted for future clarification

Agreement Status

The revised document (S4aA260009) was agreed with changes in brackets to be implemented in DaCAS-3.

Administrative Matters

IPR and Compliance

Standard 3GPP reminders provided regarding:
- IPR disclosure obligations
- Antitrust and competition law compliance
- Consensus-based decision making

Next Steps

• DaCAS meeting scheduled at SA4#135
• Submission deadline: February 3, 2026
• DaCAS-2 Editor to implement S4aA260008 revisions
• DaCAS-3 Editor to implement S4aA260009 revisions

Participation

20 participants from organizations including: Bytedance, Dolby, Ericsson, Fraunhofer IIS, Nokia, NTT, Orange, Panasonic, Qualcomm, Philips, and Xiaomi.