Ctrip Mobile Call Center Solution: Architecture, Challenges, and the CCodec Audio Codec

The authors, members of Ctrip's basic business R&D call‑center team, describe how they extended traditional call‑center infrastructure with soft‑switch, intelligent routing, and automatic speech processing to provide a human‑friendly, interactive voice service for Ctrip users.

With smartphones becoming indispensable, Ctrip developed a complete mobile seat solution that frees agents from fixed locations, allowing them to work anywhere with Wi‑Fi, 3G or 4G connectivity.

The mobile scenario retains the classic call‑center architecture but adds a Session Border Controller (SBC) at the access side for voice ingress and security control; agents can log into the seat app from a computer, smartphone or other smart device.

Mobile environments pose challenges such as unstable networks, bandwidth fluctuations, frequent handovers, latency, jitter, echo and noise, which make reliable voice communication difficult.

To address these issues, Ctrip optimized the standard SIP protocol, deeply customized an audio codec, and introduced packet‑loss compensation, reducing interaction latency and bandwidth consumption.

CCodec is a lossy audio codec built on an open‑source algorithm that supports dynamic bitrate (VBR), multiple sampling rates (8 kHz‑48 kHz), mono/stereo/multichannel, and robust packet‑loss concealment, achieving high quality even on weak networks with latency as low as 5 ms.

The codec’s performance curves show that at low bitrates it outperforms iLBC, AMR‑NB, Speex and AMR‑WB, and continues to deliver superior quality at higher bitrates.

In weak‑network scenarios CCodec can operate at 16‑32 kbps for clear voice calls, while in good conditions it can exceed 32 kbps to support high‑fidelity music streaming.

CSIP is a communication library built on SIP, SDP, RTP, STUN and integrates CCodec, offering 30% packet‑loss tolerance, 250 ms delay handling, network monitoring, dynamic bitrate adjustment, echo cancellation, and noise reduction.

Because CSIP is C‑based and complex for mobile developers, Ctrip wrapped it in PhoneSDK , providing a simple API that enables call‑out/answer, hold, transfer, mute, multi‑party conference, network status monitoring, IPv6, WebSocket, P2P, TCP/SSL/UDP, noise suppression, echo suppression, packet‑loss compensation, jitter resistance, and dynamic bitrate control.

Traffic tests show that using CCodec at 48 kHz reduces encoded audio size to about half of WeChat's and one‑third of traditional G711/PCMu under the same network conditions.

A case study with Beijing Shiyi Bei Technology Co., Ltd. demonstrates how the PhoneSDK accelerated development of a mobile seat app for insurance sales, enabling agents to handle calls and customer data on the go, with the system launched in March 2017 across multiple cities.

Recommended reading includes articles on automated testing, Java quality control, and Ctrip's fourth‑generation architecture.