A database for Amazigh speech recognition research: AMZSRD

In the field of Amazigh automatic speech recognition, speech databases are inexistent. This domain has suffered from the absence of speech database references, which allows objective comparisons between the different systems of recognition. All the published works in this field have been experimented on local databases, which contain a limited Amazigh number of digits and letters. This paper present our first version of the database that contains Amazigh speech isolated words, spoken by 50 speakers. It is designed for training and testing Amazigh speech Recognition systems. The database will be available for researches and academic uses.