Deterioration of motor neurons due to amyotrophic lateral sclerosis (ALS) would affect the strides from one gait cycle to the next. Computer-assisted techniques are useful for gait analysis, and also have high potential in quantitatively monitoring the pathological progression. In this paper, we applied the signal turns count method to measure the fluctuations in the swing-interval time series recorded from 16 healthy control subjects and 13 patients with ALS. The swing-interval turns count (SWITC) parameter derived with the threshold of 0.06 s presented a significant difference (p < 0.001) between the healthy control subjects and ALS patients. Besides the SWITC, we also computed the averaged stride interval (ASI), which is usually longer in the patient with ALS (p < 0.0001), to characterize the gait patterns of ALS patients. In the pattern classification experiments, the Fisher's linear discriminant analysis (FLDA) and the least squares support vector machine (LS-SVM), both input with the SWITC and ASI features, were evaluated using the leave-one-out cross-validation method. The results showed that the LS-SVM with sigmoid kernels was able to provide a classification accurate rate of 89.66% and an area of 0.9629 under the receiver operating characteristic (ROC) curve, which were superior to those obtained with the linear classifier in the form of FLDA.