Neurocognitive impairment is one of the core symptoms in schizophrenia and poses a great challenge to effective treatment. Sixty-one long-term hospitalized patients with schizophrenia were recruited and randomly assigned to two groups: Baduanjin exercise and brisk walking. Patients in the Baduanjin group received 24 weeks of Baduanjin training (5 days/week, 40 min/day), while patients in the brisk walking group received 24 weeks of brisk walking (5 days/week, 40 min/day). Scores on the Wechsler Memory Scale, Digit Symbol Substitution Test (DSST), and the positive and negative syndrome scale were used to evaluate the logical memory (LM), processing speed, and clinical symptoms of all participants, while the score of Trail Making Test-A (TMT-A) was applied to assess the visual attention and graphomotor speed, at baseline and the 16th week and 24th week of intervention. The one-way repeated measures analysis of variance (ANOVA) was used to test the differences in neurocognitive changes between the two groups. Repeated measures ANOVA showed significant differences between the two groups in the LM immediate (F = 6.21, p = 0.003) and LM delayed (F=5.60, p = 0.005) scores, but not in the completion times of TMT-A (F=.22, p = 0.806) or DSST scores (F=0.97, p = 0.328). A significant effect of time was also detected in the LM immediate (F=10.24, p = 0.000) and LM delayed (F=4.93, p = 0.009) scores and in the completion time of the TMT-A (F=33.10, p = 0.000), but not in the DSST scores (F=2.12, p = 0.122). Baduanjin exercise could improve logical memory in the long-term hospitalized patients with schizophrenia.