Improved Ultra-Low Bit Rate MELP Algorithm for BeiDou Satellite Communication

With the successive optimization of the BeiDou navigation satellite system (BDS), the capacity of its regional short message communication (RSMC) has increased tenfold, and the maximum length of a single short message has reached 14,000 bits, providing a theoretical basis for voice transmission. In complex usage scenarios, BDS urgently requires instant voice communication to enhance the efficiency of command and rescue. Based on the mixed excitation linear prediction (MELP) algorithm and the performance of RSMC, this paper presents an improved ultra-low bit rate MELP algorithm. The improved MELP incorporates preprocessing for distinguishing unvoiced and voiced superframes; inter-frame prediction and residual vector quantization (IFP-RVQ); and inter-frame grouping and inter-stage clustering (IFG-ISC). The aim is to increase the coding rate, save coding bits, and improve the quality of coded speech. Multiple comparative tests show that the output speech of the improved MELP has a PESQ score of 2.15, a ViSQOL score of 3.32, and a DNSMOS score of 2.99, exhibiting high intelligibility and speech quality.