Life Soundtracking: promises and challenges of automated systems aimed to recovering meaningful music for Alzheimer's patients

Title Life Soundtracking: promises and challenges of automated systems aimed to recovering meaningful music for Alzheimer's patients
Publication TypeConference Paper
Year of Publication2015
Conference NameInternational Workshop on Quantitative and Qualitative Music Therapy Research
AuthorsHerrera, P., & Gómez E.
Pagination19
Conference Start Date15/10/2015
Conference LocationBarcelona
AbstractThere is a growing amount of evidence that music that matches the likings and life experiences of Alzheimer’s patients improves their connection with caregivers and family, and increases their quality of life in general. Manually selecting the right music for a given patient is a time-consuming and error-prone critical task. Fortunately, recent advances in music information research (MIR) make it highly automatable, thanks to music analysis, similarity computation, and automatic inference. We present here a basic algorithm that, based on such techniques, creates personalized playlists for specific patients, assuming that the required information will be fragmented or even missing. The algorithm considers (after a sort of musical anamnesis) their age, places where they lived, performed activities and musical preferences and generates a playlist that expands, with likely relevant music tracks, the information provided by caregivers and patients. In addition to discussing some implementation aspects, we remark the challenges this task poses to MIR: analysis algorithms and metadata are biased towards music from the 1960’s on, the lack of annotated collections and tags that make sense to the targeted population, the special GUI and interaction flow requirements that senior users demand, and the difficulties for evaluating the systems addressed to that goal. To conclude, we elaborate on the insights provided by several evaluation rounds made with senior (but not diagnosed) users.
intranet