High-throughput mass spectrometry-based proteomics is developing rapidly in recent years. A key and essential issue in proteomics data processing is to identify proteins via tandem mass spectra. De novo peptide sequencing approach is database independent, which is a distinct advantage compared to database searching approach, so it can be used to analyze the data of new organisms or unsequenced organisms. De novo peptide sequencing problem is briefly described at first, and then the state-of-the-art of this problem is introduced from different aspects, which include the strategies with their advantages and disadvantages, frequently used algorithms and tools, criteria for algorithm assessment, and frequently used datasets for algorithm comparison. At last, the characteristics of some algorithms are summarized and some possible improvements of de novo peptide sequencing algorithm design are proposed.