Abstract: Existing soft robotic gloves face some challenges, including limited adaptability, poor portability, and unstable interactive control methods that require the use of the healthy hand, which limits their effectiveness in assisting patients with daily living activities. To address these issues, a wearable, fabric-based robotic glove with voice-bending signal fusion control is developed. The glove provides stable assistance in daily living activities through a lightweight, portable design and integrated multimodal sensing technology. A pneumatic actuator for finger bending assistance is designed using flexible materials, such as elastic bands, hook-and-loop fasteners and thermoplastic polyurethane (TPU), based on the segmented curvature of human fingers and the joint skin stretching mechanism. Experimental results demonstrate that the actuator effectively replicates the mechanical characteristics of multi-segment fingers in bending and can extend by 63.3% under a pressure of 150 kPa. Furthermore, an integrated robotic glove incorporating flexible bending and voice sensors is used to construct a wearable shoulder bag control system with integrated actuation and sensing, and a control strategy based on voice-bending signal fusion is implemented. Comparative experiments simulating daily living activities confirm the superiority of this control strategy over traditional methods, such as healthy hand control, affected-hand electromyogram (EMG) control, and single-voice control. The results indicate that the proposed strategy eliminates the need for the healthy hand, has a short system calibration time, and avoids pre-programming of assisting gestures, achieving a high average intention recognition rate. Additionally, experimental validation with real stroke patients further demonstrates that the system provides effective support for daily living activities for patients with hemiplegia due to stroke.