Design of a Phonetic Corpus for Speech Recognition in Catalan
نویسندگان
چکیده
In this paper, we present the design of a corpus for speech recognition to be used for the recording of a speech database in Catalan. A previous database in Spanish was the reference in setting the specifications about the characteristics of the sentences and in the minimum number of units required. An analysis of unit frequencies were carried out in order to know which units were relevant for training and to compare the results with the figures from the designed corpus. Three different sub-corpora were generated, one for training, the other for vocabulary-independent verification and the third for vocabulary-dependent verification. Short sentences were obtained that contained all phones and relevant diphones in a sufficient quantity. Evaluation of the corpus characteristics was performed using several parameters to validate database specifications. Using this corpus, a speech database was recorded over a telephone line and manually labelled, and it is currently used to train and test several speech recognition
منابع مشابه
Frequency analysis of phonetic units for concatenative synthesis in catalan
Knowledge of phonetic unit frequency is very necessary for developing databases in both concatenative synthesis and continuous speech recognition. In the present work, a large corpus of text was processed and phonetically transcribed to obtain allophone and diphone frequencies for the Catalan language. The corpus was acquired from newspaper articles, in which there were a lot of foreign words t...
متن کاملAllophone-based acoustic modeling for Persian phoneme recognition
Phoneme recognition is one of the fundamental phases of automatic speech recognition. Coarticulation which refers to the integration of sounds, is one of the important obstacles in phoneme recognition. In other words, each phone is influenced and changed by the characteristics of its neighbor phones, and coarticulation is responsible for most of these changes. The idea of modeling the effects o...
متن کاملTaking Advantage of Spanish Speech Resources to Improve Catalan Acoustic HMMs
At TALP, we are working on speech recognition of official languages in Catalonia, i.e. Spanish and Catalan. These two languages share approximately 80 % of their allophones. The speech databases that we have available to train HMMs in Catalan have a smaller size than the Spanish databases. This difference of size of training databases results in poorer phonetic unit models for Catalan than for ...
متن کاملVOXMEX Speech Database: Design of a Phonetically Balanced Corpus
We present a method for designing a phonetically balanced speech corpus. In this method, we used a phonotactic approach to design the phonetic content of VOXMEX: a phonetically balanced corpus for Mexican Spanish. The transcriptions of VOXMEX contain a complete coverage of phonemes and allophones of Mexican Spanish in every possible context. This corpus is designed for doing phonetic research a...
متن کاملDesign and analysis of a German telephone speech database for phoneme based training
Based on the Sotscheck text corpus, we developped a new corpus that was specifically optimised for training phoneme-based recognition systems. Particular attention was payed on good coverage of phone transitions. Even though the resulting corpus is only slightly enlarged, it shows an increased phonetic coverage while maintaining a good phonetic balance. Results of phonetic statistical analysis ...
متن کامل