Sputum is the most common sample collected from patients suffering from lower respiratory tract infections and it is crucial for the bacterial identification of these infections. In this study, we enrolled 101 sputum samples from 101 patients with lower respiratory tract infections. Initially, pyrosequencing of the 16S rDNA V3 hypervariable regions of the bacteria contained in the sputum was utilized as a culture-independent approach for microbiota analysis. For comparison, clinical laboratory tests using a culture-dependent automated bacterial identification system for the same cohort of sputum samples were also done. By pyrosequencing, >70,000 DNA fragments were found and classified into 129 bacterial genera after being analyzed by the Ribosomal Database Project (RDP) process. Most sequences belonged to several predominant genera, such as Streptococcus and Staphylococcus, indicating that these genera play an important role in lower respiratory tract infections. In addition, some sequences belonging to potential causative agents, such as Mycoplasma, Haemophilus, and Moraxella, were also found, but these sequences were not found by clinical laboratory tests. For the nine genera detected by both methods, the methods' sensitivities were compared and the results showed that pyrosequencing was more sensitive, except for Klebsiella and Mycobacterium. Significantly, this method revealed much more complicated bacterial communities and it showed a promising ability for the detection of bacteria.