All participants will receive trainings for 90 minutes per day, three days per week for 12 weeks (a total of 36 training sessions). The participants in the DUAL group will be instructed to perform physical exercise and cognitive tasks simultaneously (e.g., math calculation while stepping). The difficulty of the cognitive tasks will increase as the participants improve in their performance. The 90 minutes of training session will be break up into 2 to 3 parts, and the participants can rest as needed.

The physical exercise programs will involve balance or strength training or aerobic exercises. These may include (but not limited to) stepping, walking, dancing, ball kicking and throwing, and etc. The exercise session will be break up into 2 to 3 parts, and the participants can rest as needed during the exercise period. The exercise intensity will be progressed as the participants improve their performance throughout practice. To prevent exercise-induced injuries, vital signs and the Borg Perceived Exertion Scale will be monitored and recorded in each training session. If the training therapist observed that the exercise may be too intense for the participants, he/she will immediately reduce the training intensity.

The WAIS-III is developed to measure an individual's intelligence level. It includes tests that evaluate cognitive functions in verbal comprehension, working memory, perceptual organization, and processing speed. The subtests that we will use are the Digit Symbol-Coding and Matrix Reasoning tests. The Digit Symbol-Coding test consists of 9 digit-symbol pairs and the participants will be asked to write down the corresponding symbols for the given digits on the test sheet as accurately and as fast as possible. As for the Matrix Reasoning test, the participants will need to logically solve missing puzzles within given matrixes. The test result of the Matrix Reasoning test indicates a general intelligence level because it entails the abilities of visual-spatial reasoning, abstract reasoning, visual organization, and visuospatial information processing.

The UFOV assessment is a computer-based visual test containing three subtests: visuomotor processing speed, divided attention, and selective attention (Ball, Edwards, & Ross, 2007). The UFOV is the visual area over which information can be extracted from a brief glance without eye or head movements.

The Stroop Color-Word assesses the abilities of selective attention, inhibition and executive function. The participants will be tested under congruent and incongruent conditions. In the congruent condition, the participant will name the color ink of a word which is consistent with the written color name; whereas in the incongruent condition the participant will name the color ink differs from the written color name. The time to complete the task will be calculated for each condition (Koss, Ober, Delis, & Friedland, 1984; Ridley, Johnson, & Braisted, 1978).

The dual-task test evaluates the ability to shift attention between one task and another. Participants will sit and perform the box and block test (BBT) or walk 10 meters while doing secondary cognitive or motor tasks. Two cognitive secondary tasks will be performed by the participants: (1) arithmetic task: participants will be asked to perform serial subtractions by 7 starting from 100 or random two-digit numbers (e.g., Baetens et al., 2013); (2) tone discrimination task: participants will be presented a number of low and high-pitched tones and they will respond to either the high or low-pitched tones during the trial. Both cognitive task performances will be recorded and the results will be compared to single cognitive task performance. In addition to the cognitive dual-task, participants will perform a motor task (e.g., holding a cup of water) while walking.

The TUG assesses the dynamic balance ability and mobility. The participants will be required to stand up from a chair, walk 3 meters, turn around, walk back to the chair, and sit down. The time to complete the TUG test has been shown to be a good indicator to detect potential fallers in frail elderly (Podsiadlo & Richardson, 1991). The test-retest reliability of TUG on individuals with cognitive impairment was excellent (Blankevoort, van Heuvelen, & Scherder, 2013).

The 30 second sit-to-stand will be assessed to indicate the strength and endurance level of the lower extremities. The participants will be asked to stand up from a standardized chair and then sit down as many times as possible within 30 seconds. The feasibility and reliability of using sit-to-stand test in people with cognitive impairment have been established to be good (Blankevoort et al., 2013).

The Chinese version of the International Physical Activity Questionnaires (IPAQ) is an international measure for health-related physical activity. A short form of IPAQ will be used to assess changes in physical activity before and after intervention in this study. The reliability and validity of IPAQ has been established across 12 countries (Craig et al., 2003).

Accelerometers will be used to provide an objective measure of the amount of arm movements in real-life situations. The participants will be asked to wear an Actigraphy activity monitor (ActiGraph, Shalimar, FL, USA) on both wrists for 3 consecutive days before and after training to measure the number of moves each minute, and the average counts of move per minute. The participants will be required to wear the device during the day except for doing water-based activities, such as bathing or swimming. Data recorded by the actigraphy will be analyzed with the MAHUFFE software (http://www.mrc-epid.cam.ac.uk/).

We will evaluate isometric knee flexors and extensors muscle strength using handheld dynamometer. The participant will be seated upright in a chair with back support, the knee will be placed in 90-degree flexion and the evaluator will stabilize the thing to eliminate synergistic movements. Participants will be asked to perform a maximal isometric contraction of knee flexion and extension with affected and less affected side. We will record the mean value of 3 attempts.

We will use hand dynamometer to measure grip strength of both hands while the participant is seated, with the elbow at 90-degree flexion. We will record the mean value of 3 attempts.

Assess activities of daily living

The Lawton IADL scale evaluates 8 activities with a score range from 0 to 8 (higher indicate better function). The inter-rater reliability and validity of the Lawton IADL have been established to be moderate to high for community-dwelling older adults (Graf, 2008; Lawton & Brody, 1969).

Assess activities of daily living.

The intervention effects on quality of life will be assessed with the Chinese versions of the Quality of Life in Alzheimer's Disease Instrument (QoLAD)

The social participation level will be assessed with the CIQ. The CIQ measures 15 items relevant to home integration, social integration, and productive activities (Willer, Ottenbacher, & Coad, 1994).The CIQ has been tested on various populations with acquired brain injuries, and test-retest reliability was excellent for chronic stroke patients with aphasia (Dalemans, de Witte, Beurskens, van den Heuvel, & Wade, 2010; Willer et al., 1994).

The Chinese version of CBI will be used.

The Geriatric Depression Scale (GDS) - 15 items version is a self-administered questionnaire used to evaluate mood and depressive symptoms. The scores range is 0-15 and a score of 5 or greater taken as a possible indicator of depression.