Information coding frame loss recovery method and device

Abstract

The invention discloses a method and a device for recovering information coding frame loss. The method includes: receiving a plurality of information coding frames and redundant frames, and the redundant frames include at least one information coding frame extracted from the plurality of information coding frames according to a preset interval. frame data; detect the missing information encoding frame before the current information encoding frame; when the redundant frame does not contain the missing information encoding frame, obtain the data of the adjacent frames of the missing information encoding frame from the current information encoding frame and the redundant frame ; According to the data of the adjacent frame, recover the data of the missing information coding frame. In the case of the same redundant bandwidth, the present invention has more than doubled the capability of repairing lost packets compared with the traditional FEC method, and has obvious improvement on the subjective sense of hearing.

Description

technical field [0001] 本发明涉及互联网技术领域,尤其涉及一种信息编码帧丢失恢复方法和装置。 Background technique [0002] Voip(Voice over Internet Protocol,指在IP网络上使用IP协议以数据包的方式传输语音)在实际应用中为了保证语音传输的实时性,通常使用UDP(User DatagramProtocol的简称,中文名是用户数据报协议)不可靠连接协议方式,而互联网普遍存在带宽竞争、传输拥塞现象,导致数据包极容易发生丢失,而影响语音通话质量最大的因素就是数据包丢失。解决网络丢包问题,目前主要有三种方法: [0003] 1、PLC(丢包隐藏) [0004] 常规的PLC是在解码端进行丢包补偿处理,其基本原理是利用前一帧或邻接帧正常数据包的解码信息,采取信息预测方法尽可能地恢复丢失的语音帧信息。PLC方法的缺点:在连续丢帧的情况下,由于PLC所补偿的各个语音帧具有相同的频谱特性(相同的LPC造成)和基音频率,容易引入一种可察觉的噪声,尤其是当基音频率较高的时候,这种因为过分的周期性所引起的。所以在连续丢帧的情况下,为了减少各个补偿帧之间的相关性,PLC会将能量进行逐帧递减,以减轻不自然噪声感,同时丢包恢复能力减弱。因此,常规的PLC主要针对1-2个连续丢包比较有效。 [0005] 2、FEC(前向纠错校正) [0006] FEC是在发送端发出的音频数据中增加冗余信息,接收端根据冗余信息检测和纠正丢包信息,一般而言FEC编码的冗余度越大其纠错能力越强,但冗余度越大,意味着冗余数据占用的带宽越多,带宽利用率越低。 [0007] 3.ARQ(自动重传请求) [0008] 由于Voip使用UDP方式,而UDP没有重传机制,而ARQ就是在网络发生丢包时自动重传丢失的数据包,接收端需要缓冲和排序已收到的数据包,IP网络时延和抖动等因素对ARQ性能有很大影响,ARQ在判断丢包、发送丢包请求、重传包重新发送等环节都引入时延,所以容易导致通话时延增加,影响通话主观体验。 [0009] 目前,大部分语音解码器会自带有PLC(丢包隐藏)功能,但只能有限度地修复少量丢包(例如1-2个包),无法恢复连续多个丢包,如果连续丢包在3个以上,解码器的PLC功能基...

AI Technical Summary

Problems solved by technology

Disadvantages of the PLC method: In the case of continuous frame loss, since each speech frame compensated by PLC has the same spectral characteristics (caused by the same LPC) and pitch frequency, it is easy to introduce a perceivable noise, especially when the pitch frequency When it is high, this kind of excessive periodicity caused by

[0008] Since Voip uses UDP, and UDP has no retransmission mechanism, and ARQ is to automatically retransmit the lost data packets when the network loses packets, the receiving end needs to buffer and sort the received data packets, IP network delay and jitter, etc. Factors have a great impact on ARQ performance. ARQ introduces delays in judging packet loss, sending packet loss requests, and retransmitting packets. Therefore, it is easy to increase the call delay and affect the subjective experience of the call.

[0009] At present, most voice decoders will have PLC (packet loss concealment) function, but it can only repair a small amount of packet loss (for example, 1-2 packets) to a limited extent, and cannot recover multiple consecutive packet loss. When there are more than 3, the PLC function of the decoder is basically invalid, and the packet loss recovery ability is not strong, and even obvious unnatural noise can be heard. In addition, the human ear can clearly perceive the freeze caused by packet loss under a large number of continuous packet loss. and other phenomena, which greatly affect the subjective experience of the call; some decoders use FEC forward error correction correction technology when repairing packet loss, but as the number of consecutive frames to be repaired increases, the additional bandwidth required by FEC will increase exponentially, otherwise the repair effect It is not good, so there are certain disadvantages; while the decoder adopts ARQ, it will introduce delay, which is not suitable in the network environment with a lot of packet loss

Images