The design of a hands-free speech recognition application during the intrapartum stage

Abstract

Unlike the developed nations, the health sector within the developing countries is faced with the triple challenges of human, financial and technological scarcity of resources. This insufficiency of resources results into amongst other intrapartum mishaps. To ameliorate some of these conditions, the World Health Organization (1994) promoted the use of the pathogram as an informative and data capturing tool that could help reduce intrapartum mishaps. The usage of the partogram within the intrapartum environment also introduced a dilemma as birth attendants spent quite a good amount of time using their eyes and hands (in pen and paper) capturing medical data onto the partogram instead of investing these resources onto the expectant mother and or fetus. This study adopted Design Science Research as a suitable research approach, strengthened by a pragmatic philosophical standpoint. This study involved the following methods; • A review of literature in the intrapartum environment, along with topics from relevant reference disciplines including speech recognition • A series of semi-structured contextual interviews with birth attendants, student nurses and senior midwives • A design science research study using the knowledge from the reference disciplines to design a hands-free voice driven epartogram • A simulation of the capturing of intrapartum data to evaluate and refine the prototype (epartogram) by applying anonymized intrapartum data driven by natural speech • An evaluation of the artifact (epartogram) based on a number of published guidelines recommended by scholars to demonstrate its potential utility as well as to establish if the solution is generic to the contextual environment. Although the introduction of ICT into the problem domain abetted the process of data capturing (specifically the referral process), the fundamental aspect of using the prototype to free the hands and eyes of the birth attendants proved challenging due to issues with the recognition of natural speech by speech recognition systems and background noise. Monitoring of MOU and the referral process from a lower MOU to a higher one could benefit a great deal from this study as the prototype thrived well in that regard. Natural speech recognition by machines in an uncontrolled environment is still at its infancy (some of the most powerful engines can not differentiate between background noise and direct instruction). Not withsatnding the challenges posed by the infancy of speech recognition, the artifact showed potential as a manual epartogram providing spatial access to multiple participating MOU via the cloud.