Dementia poses significant challenges to individuals, particularly those grappling with Alzheimer's disease, impacting their daily lives, well-being, and communication. The task of enhancing communication for People with Dementia (PwD) has grown increasingly complex, prompting the exploration of innovative solutions such as voice assistants. Understanding PwD's speech features and conversational content can enhance their communication experiences. Speech feature analysis plays a pivotal role in this process, offering valuable insights into tailoring interactions with PwD. In this study, we analyze speech recordings from the ADReSSo database, which contains speech from both Alzheimer's patients and healthy aging persons. We discovered that several speech features exhibited significant differences between PwD and their healthy counterparts. These speech features can help with dementia identification and also represent the PwD's speaking patterns. Furthermore, the speech emotions of PwD and healthy aging people differ significantly. The clear differences in speech emotions offer valuable cues for crafting emotionally attuned interactions between PwD and voice assistants. These valuable information serves as the foundation for developing future personalized voice assistants dedicated to supporting individuals with dementia by incorporating the identified speech features and emotional information.