Human emotions are essential to recognize the behaviour and the state of mind of a person. Emotion detection through speech signals has started to receive more attention lately. Living alone could be hard for some people due to the lack of social interaction, as they might develop a series of negative emotions daily. Furthermore, there are some unavoidable circumstances when family members need to live away from their families, leaving their old parents to live alone. These circumstances may cause parents to experience anxiety or a decline in mental health, which is a major cause for concern for their children. This is where assisted living technology can come in to support. This research proposes the design and development of a speech emotion recognition system for solitary people to detect and monitor their state of mind as well as their daily emotional behaviour. The research has three main contributions. First, to implement a real-time system based on audio where we can predict emotions from recorded human voices via deep learning. Secondly, a model has been designed to use data normalization and data augmentation techniques for advanced classification. Finally, a speech emotion detection system has been created using a Long Short Term Memory (LSTM) recurrent neural network. This research aims to study solitary person activities at any time at home. The resulting system will be used for mental health monitoring.