Accepted Silence Insertion Descriptor frame

Description

An Accepted SID (Silence Insertion Descriptor) frame is a concept used in speech codecs, particularly in the context of Discontinuous Transmission (DTX) and comfort noise generation. In 3GPP specifications, it is defined as a traffic frame in which the SID field deviates in less than 16 bit positions from the SID code word (with flag SID=2 or SID=1). This definition is found in specifications such as TS 46.031, TS 46.041, and TS 46.081, which pertain to the Adaptive Multi-Rate (AMR) and other speech codecs.

During a voice call, when the speaker is silent, the transmitter may stop sending speech frames to save bandwidth and power. Instead, it sends Silence Insertion Descriptor (SID) frames that contain parameters for generating comfort noise at the receiver. The receiver uses these parameters to synthesize background noise, making the call more natural and avoiding the impression of a dropped connection. The SID frames are transmitted at a lower rate than speech frames, typically every 160 ms or so.

The SID field within a traffic frame is a specific bit pattern that identifies the frame as a SID frame. The SID code word is a predefined sequence of bits. Due to transmission errors, the received SID field may not exactly match the code word. The standard specifies that if the SID field deviates in less than 16 bit positions from the SID code word, the frame is considered an 'Accepted SID frame'. This tolerance allows for some bit errors while still correctly identifying the frame as a SID frame. The flag SID=2 or SID=1 indicates the type of SID frame (e.g., SID=2 for the first SID frame after speech, SID=1 for subsequent ones).

The concept of an Accepted SID frame is important for robust operation in the presence of transmission errors. It ensures that the receiver can correctly detect SID frames even if some bits are corrupted, thus maintaining comfort noise generation and avoiding unnecessary muting or artifacts. The threshold of 16 bit positions is chosen to balance between false positives (accepting a non-SID frame as SID) and false negatives (rejecting a valid SID frame due to errors).

In the AMR codec, the SID frame is part of the comfort noise generation mechanism. The AMR speech codec operates at various bit rates, and during DTX, it uses SID frames to update the comfort noise parameters. The receiver's comfort noise generator uses these parameters to produce background noise that matches the original noise characteristics. The Accepted SID frame definition ensures that the receiver can reliably detect these frames and update the comfort noise accordingly.

The evolution of this term has been stable since its introduction in Release 8. The definition has remained unchanged across releases, indicating that the technical requirements for SID frame detection have not changed. This stability reflects the maturity of the speech codec specifications and the robustness of the existing mechanisms.

Purpose & Motivation

The Accepted SID frame concept exists to ensure reliable detection of Silence Insertion Descriptor frames in the presence of transmission errors. In voice communication, DTX is used to reduce bandwidth during silence periods by sending SID frames instead of regular speech frames. These SID frames contain parameters for comfort noise generation at the receiver. However, transmission errors can corrupt the bits of the SID field, potentially causing the receiver to misinterpret the frame. The Accepted SID frame definition provides a tolerance: if the received SID field is within 16 bit positions of the SID code word, it is accepted as a valid SID frame. This improves robustness and ensures that comfort noise is generated correctly even with some bit errors.

Without this concept, the receiver might discard SID frames that have minor errors, leading to gaps in comfort noise and a degraded user experience. Alternatively, it might accept frames that are not SID frames, causing incorrect noise generation. The 16-bit threshold is a compromise that has been found to work well in practice. It addresses the limitations of previous approaches that might have required exact match, which is impractical in real-world error-prone channels.

The concept is part of the broader effort to optimize voice transmission over cellular networks, balancing quality and efficiency. It was introduced in Release 8 as part of the specifications for the AMR codec and has been carried forward unchanged, demonstrating its effectiveness and the stability of the underlying technology.