Welcome to KingLine Data Center!   Contact Phone: 0086-10-62660053   Email: marketing@speechocean.com

English

Speechocean HomepageHelp

Home > Commercial Resources

Recording Platform

>

All

Language

>

All

Sort By:Default

27 Results

King-AVT-002

This database contains 100 hours of speech which was transcribed and annotated from home videos collected from real life of Chinese native speakers.

King-TTS-024

The Japanese Speech Synthesis Corpus contains the recordings of 1 voice talent (male, professional broadcaster, 30 years old, born and raised in Tokyo). It is recorded in a professional studio over two channels (waveform and electroglottography (EGG) signal). The script contains 8864 (approx.) utterances in total. There are 26 categories in this corpus, which contains the most common language phenomenon in Japanese. Speech rate consistency, energy consistency and tone color consistency are strictly controlled during recording process. Each utterance wave is stored in Windows uncompressed PCM format. Manual Prosody labeling, Phone Boundary Labeling were adopted based on the real speech. A pronunciation lexicon is available with a phone set in PLU. All manually checked. Each utterance was carefully proofread by linguists, to make sure the phonetic symbol is consistent with the real pronunciation. Misread utterances have been re-recorded.

King-TTS-026

The Hong Kong Cantonese Speech Synthesis Corpus contains the recordings of 1 voice talent (female, professional broadcaster, 30 years old, born and raised in Hong Kong). It is recorded in a professional studio over two channels (Shure SM15 microphone and Electroglottography (EGG) sensor). The script contains 8936 (approx.) utterances in total. There are 27 categories in this corpus, which contains the most common language phenomenon in Hong Kong Cantonese. Speech rate consistency, energy consistency and tone color consistency are strictly controlled during recording process. Each utterance wave is stored in Windows uncompressed PCM format. Manual Prosody labeling, Phonetic Tone Labeling, Phone Boundary Labeling were adopted based on the real speech. A pronunciation lexicon is available with a phone set in Cantonese Pinyin. All manually checked. Each utterance was carefully proofread by linguists, to make sure the phonetic symbol is consistent with the real pronunciation. Misread utterances have been re-recorded.

King-TTS-027

The Hong Kong Cantonese Speech Synthesis Corpus contains the recordings of 1 voice talent (male, professional broadcaster, 30 years old, born and raised in Hong Kong). It is recorded in a professional studio over two channels (Shure SM15 microphone and Electroglottography (EGG) sensor). The script contains 9004 (approx.) utterances in total. There are 28 categories in this corpus, which contains the most common language phenomenon in Hong Kong Cantonese. Speech rate consistency, energy consistency and tone color consistency are strictly controlled during recording process. Each utterance wave is stored in Windows uncompressed PCM format. Manual Prosody labeling, Phonetic Tone Labeling, Phone Boundary Labeling were adopted based on the real speech. A pronunciation lexicon is available with a phone set in Cantonese Pinyin. All manually checked. Each utterance was carefully proofread by linguists, to make sure the phonetic symbol is consistent with the real pronunciation. Misread utterances have been re-recorded.

King-TTS-030

The British English Speech Synthesis Corpus contains the recordings of 1 voice talent (female, professional broadcaster, 34 years old, born and raised in London). It is recorded in a professional studio over two channels (Shure SM15 microphone and Electroglottography (EGG) sensor). The script contains 8810 (approx.) utterances in total. There are 20 categories in this corpus, which contains the most common language phenomenon in British English. Speech rate consistency, energy consistency and tone color consistency are strictly controlled during recording process. Each utterance wave is stored in Windows uncompressed PCM format. Manual Prosody labeling, Stress Labeling, Phone Boundary Labeling were adopted based on the real speech. A pronunciation lexicon is available with a phone set in BEEP. All manually checked. Each utterance was carefully proofread by linguists, to make sure the phonetic symbol is consistent with the real pronunciation. Misread utterances have been re-recorded.

King-NLP-034

This corpus contains 220,000 US English SMS sentences. All the words were processed by filtering out the sensitive words, noisy words, repeated chatting sentences. All text sentences were classified into different categories and were proofreaded manually.

King-NLP-037

The corpus contains 74,300 sentences, 2,835,290 Chinese characters. It was collected from the daily life and business of individual person from Taiwan and Hong Kong by getting his/her authorization. All words were proofreaded manually.<br /> *Only for domestic market.

King-ASR-044

The Taiwan Mandarin Mobile Speech Recognition Corpus was collected in Taiwan. It contains the voices of &nbsp;5232 different speakers (2365 males, 2867 females) who were balanced distributed in age (mainly 16 &ndash; 30,31 &ndash; 45,46 &ndash; 60), gender and regional accents (for the details, please see the technical document). The script contains 1,643,521(approx.) utterances in total (for more details of script structure design, please check the specification), specially designed to provide materials for both training and testing of many classes of speech recognizers. Each speaker was recorded in 2 environments: quiet enviroments (office, home) and noisy enviroments (cafe, restaurant, street). Mobile platform, i.e. Android was used for speech collection. Each utterance wave was stored in a separate file and uncompressed. A pronunciation lexicon is available with a phonemic transcription in Pinyin. All manually checked. All audio files were manually transcribed and annotated by native transcribers. Details are available with specification.

King-ASR-080

The Canada French Mobile Speech Recognition Corpus was collected in Canada. It contains the voices of 50 different speakers (25 males, 25 females) who were balanced distributed in age (mainly 16 &ndash; 30,31 &ndash; 45,46 &ndash; 60), gender and regional accents (for the details, please see the technical document). The script contains 25,049(approx.) utterances in total, covering 16 categories and 44 sub-categories(for more details of script structure design, please check the specification), specially designed to provide materials for both training and testing of many classes of speech recognizers. Each speaker was recorded in a quiet office environment. Mobile platforms, i.e. iOS\Android\Windows were used for speech collection. Each utterance wave was stored in a separate file and uncompressed. A pronunciation lexicon is available with a phonemic transcription in XSAMPA. All manually checked. All audio files were manually transcribed and annotated by native transcribers. Details are available with specification.

King-ASR-125

The Japanese In-Car Speech Recognition Database is collected in Japan. It contains the voices of 309 different native speakers (134males, 175 females) who were balanced distributed in age (mainly 16 &ndash; 30,31 &ndash; 45,46 &ndash; 65), gender and regional accents (for the details, please see the technical document). The script contains 98,981(approx.) utterances in total, 333.3 utterances for each speakers, covering 18 categories and 55 sub-categories (for more details of script structure design,please check the specification), specially designed to provide material for both training and testing of many classes of speech recognizers. Each speaker was recorded in 2 environments out of 7 optional environments (STOP_MOTOR_RUNNING, LOW_SPEED_ROUGH_ROAD, &nbsp;HIGH_SPEED_GOOD_ROAD, etc) with various recording conditions (motor running, fan on/off, window up/down and etc). 4 high quality audio channels Shure SM10A / Sennheiser ME104 /AKG C400BL/AKG C400BL were used in a car, at least 3 popular car models were adopted. The recording software (AUDIOREC) is developed by Speechocean. Each utterance wave was stored in a separate file as uncompressed PCM raw data. Each file is accompanied by an ASCII SAM label file which contains the relevant descriptive information. All audio files were manually transcribed and annotated by native transcribers. Details are available with Specification. The corpus follows the general convention of SpeechDat-Car.

1 2 3