niger-volta-lti / urhobo-asr-spoken-digits

URH-DIGITS is a connected digits speech recognition task

Urhobo Spoken Digits

• Item Name: URH-DIGITS
• Author(s): {I, JN} Orife
• Data Source: lavalier microphone
• Audio Format: 1-channel Linear PCM, 16k, 16bit
• Application: Speech Recognition
• Language: Urhobo
• Language ID: urh

URH-DIGITS contains speech collected for the purpose of bootstrapping Urhobo ASR modeling efforts with the task of recognizing connected digit sequences. There is currently a single speakers pronouncing 150 digit sequences.

The corpus was collected in an open acoustic environment with a lavalier microphone, digitized at 16kHz. The waveform files are in linear PCM format. All audio files were manually transcribed and annotated by native speakers.

URH-DIGITS is modeling after TIDIGITS, an English language connected digits recognition task

Resources:

• http://white.ucc.asn.au/Kaldi-Notes/tidigits/
• http://www.ux.uis.no/norsig/norsig2003/Proceedings/papers/cr1027.pdf
• https://groups.google.com/forum/#!topic/kaldi-help/tzyCwt7zgMQ
• https://github.com/Jakobovski/free-spoken-digit-dataset

Papers:

• Performance Analysis of Spoken Arabic Digits Recognition Techniques
• Pashto Spoken Digits
• Spoken Digits Recognition