RWCP Sound Scene Database in Real Acoustical Environments Non-speech sound dry source |
[non-speech top]
[sounds]
[sample]
[research 1]
[research 2]
[system]
Kazuo Hiyane and Jun Iio, Mitsubishi Research Insitute, Inc.
RWCP Sound Scene Database in Real Acoustical Environments Non-speech sound dry source |
[non-speech top]
[sounds]
[sample]
[research 1]
[research 2]
[system]
Kazuo Hiyane and Jun Iio, Mitsubishi Research Insitute, Inc.
| 1. Recording policy |
 |
In a real environment, there are many kinds of sounds other than speech. In contrast to speech recognition, until now non-speech sounds have seldom been studied. "Non-speech sound dry source database" has been developed to contribute to studies on non-speech sound recognition and removal of background sound in speech recognition.
It is impossible to record all kinds of non-speech sounds. Even with same sound source, various waveforms can be produced according to the acoustic characteristics of a room. Hence, in the database, recording was conducted in an anechoic room as a dry source, which contains a sound source wave not influenced by the acoustic characteristics of the room. In addition, each sound source was experimented by changing beating and ringing manners to yield many samples (each sample size was 100 per sound source) necessary for study on sound source discrimination.
Convolution of the dry source with the impulse response of each kind of the room allows the reproduction of a sound in the room, and thus diversity of sounds, necessary for the development of a reliable algorithm for sound source discrimination.
| 2. Outlines of recording |
|
Using a standard single microphone, acoustic signals of about 100 types of sound sources were measured in an anechoic room as the dry source.
| Recording location | unechoic room |
|---|---|
| Recording equipment | Standard microphone (B&K 4134), Filter amp. (B&K 2636), DAT Recorder (SONY DTC-77ES) |
| 48kHz, 16bit | |
| Number of samples | 9722 |
| Amount of data | 824MB (16bit RAW format) |
| SNR | 40〜50dB |
The acoustic signals recorded are classified in three categories according to the types of sound sources as shown in the following table.
| Category | # of samples | Sound source [Japanese onomatopoeia] | |
|---|---|---|---|
| Collision sound source | a1.Collision sound source (wood) | 1187 | wood board,wood stick [kon, poku] |
| a2.Collision sound source (metal) | 1000 | metal board, metal can [kan, kin] | |
| a3.Collision sound source (plastic) | 550 | plastic case [kasha, karara] | |
| a4.Collision sound source (ceramic) | 800 | glasss, china [chin] | |
| Action sound source | b1.article dropping | 200 | dropping articles in box [Za:, barabara] |
| b2.gas jetting | 200 | spray, pump [shu:] | |
| b3.rubbing | 500 | sawing, sanding [gi:] | |
| b4.bursting and breaking | 200 | breaking stick, air cap [baki, puchi] | |
| b5.clapping sound | 829 | hand clap, slamming clip [pan, pon] | |
| Characteristic sound source | c1.small metal articles | 1072 | small bell, coin [chirin, shan] |
| c2.paper | 400 | droping book, tearing paper [basa, bee] | |
| c3.musical instruments | 1079 | drum, whistle, bugle [pon, pee, pafu] | |
| c4.electronic sound | 705 | phone, toy [pipi, buu] | |
| c5.mechanical | 1000 | spring, stapler [jii, gacha] |
| 3. Recording conditions |
|
Recording environment was the anechoic room. Specification of the anechoic room was described below. Actual noises measured level were 17.3 dBA and 44.3 dBC.
| Category | Instrument | Maker | Model |
|---|---|---|---|
| Receicer | Microphone | B&K | 4134 1/2inch Measuring Microphone |
| Amplification | Microphone amplifier filter | B&K | 2636 Measuring Amp. |
| Recording | DAT recorder Sony | Sony | DTC-77ES |
The microphone was fixed 20 cm from the floor (piano wire network), while the sound source was set 5 - 20 cm from the floor. The sound was generated from the source by such methods as beating the object while handholding it, beating the object while it was placed on a sound absorbing board on the floor and dropping the object on a board on the floor.
The distance between the sound source and the microphone was basically set at 1.0 m. To gain a higher S/N ratio, according to sound volume, two gains of the microphone amplifier were set to +30 dB and +40 dB. In case of very small sound, +40 dB was selected and the distance was decreased to 50 cm. In the microphone amplifier, a 22.4 kHz low path filter was used.
|
|
| Steps of editing data recorded |
|
See calculation method for low path filter and down sampling
See convolution calculation method of the dry source with the impulse response
| List of data recorded |
|
The following data is included in the DVD-ROM.
The spectrogram (progression of power spectrum over time), graph image and data for test listening included in the 92 types of sound sources are each only a representative one.
In the signal data reconstructed by convoluting the impulse response with the dry source, 0.005 raw of 48 kHz of respective sound source was used. In other words, each sound source contains one data.
The structure of a directory is stratified for each sound source in the dry source and for the impulse response in the reconstructed data.
/nospeech … Current directory +---/drysrc ………… Dry source +---/a1 ………… Collision sound source (wood) | +---/cherry1 Beating handheld wooden board (small) B with a wooden stick | | +--- index.htm | | +--- cherry1.gif Graph image of waveform | | +--- cherry1.jpg Graph image of spectrogram | | +--- cherry1.spg Spectrogram data | | +--- cherry1.wav Sound for replay | | +---/48khz | | | +--- 002.raw Dry source (48kHz) | | | : ： | | | +--- 100.raw Dry source(48kHz) | | | | | +---/16khz | | +--- 002.raw Dry source(16kHz) | | : ： | | +--- 100.raw Dry source(16kHz) | | | +---/cherry2 Beating handheld wooden board (small) B with a wooden stick | +---/cherry3 Beating handheld wooden board (medium) B with a wooden stick | +---/magno1 Beating handheld wooden board (small) C with a wooden stick | +---/magno2 Beating handheld wooden board (medium) C with a wooden stick | +---/magno3 Beating handheld wooden board (large) C with a wooden stick | +---/teak1 Beating handheld wooden board (small) A with a wooden stick | +---/teak2 Beating handheld wooden board (medium) A with a wooden stick | +---/teak3 Beating handheld wooden board (large) A with a wooden stick | +---/wood1 Beating handheld wooden board (small) A and wooden board (small) B against each other | +---/wood2 Beating handheld wooden board (medium) A and wooden board (medium) B against each other | +---/wood3 Beating handheld wooden board (large) A and wooden board (large) B against each other | +---/a2 ………… Collision sound source (metal) | +---/bank Beating a handheld moneybox with a metal stick | +---/bowl Beating a handheld ball with a metal stick | +---/candybwl Beating a handheld metal box with a metal stick | +---/coffcan Beating a handheld coffee can with a metal stick | +---/colacan Beating a handheld Coca-Cola can with a metal stick | +---/metal15 Beating a handheld metal board (1.5 mm in thickness) with a metal stick | +---/metal10 Beating a handheld metal board (1.0 mm in thickness) with a metal stick | +---/metal05 Beating a handheld metal board (0.5 mm in thickness) with a metal stick | +---/pan Beating a handheld pan with a metal stick | +---/trashbox Beating a handheld dustbox with a metal stick | +---/a3 ………… Collision sound source (plastic) | +---/case1 Beating handheld plastic cases A, B and C with a wooden stick | +---/case2 Beating handheld plastic cases A, B and C against each other | +---/case3 Dropping plastic cases A, B and C on a plywood board | +---/dice1 Dropping dice A, B, C, D and E on a wooden board (large) | +---/dice2 Dropping dice A, B, C, D and E on a plywood board | +---/dice3 Dropping dice A, B, C, D and E on a metal board | +---/a4 ………… Collision sound source (ceramic) | +---/cup1 Beating glass cups A, B, C, D and E placed on a sound absorbing board with a wooden stick or a spoon | +---/cup2 Beating handheld glass cups A, B, C, D and E with a wooden stick or a spoon | +---/bottle1 Beating side downward of glass bottles A, B, C, D and E placed on a sound absorbing board with a wooden stick or a spoon. | +---/bottle2 Beating mouth downward of glass bottles A, B, C, D and E placed on a sound absorbing board with a wooden stick or a spoon. | +---/china1 Beating side downward of china A, B, C, D and E placed on a sound absorbing board with a wooden stick or a spoon. | +---/china2 Beating bottom downward of china A, B, C, D and E placed on a sound absorbing board with a wooden stick or a spoon | +---/china3 Beating side downward of handheld china A, B, C, D and E with a wooden stick or a spoon | +---/china4 Putting china A, B, C, D and E one of top of the other on a sound absorbing board | +---/b1 ………… Action sound source (article dropping) | +---/particl1 Dropping articles A, B, C and D in a paper box | +---/particl2 Dropping articles A, B, C and D in a metal box | +---/b2 ………… Action sound source (gas jetting) | +---/spray Sound of gas spray A and B | +---/pump Sound of air pump | +---/b3 ………… Action sound source (rubbing) | +---/saw1 Sawing a metal piece with a metal saw | +---/saw2 Sawing a wood piece with a jigsaw | +---/file Filing a metal stick with a metal file | +---/sandpp1 Sanding a wooden piece | +---/sandpp2 Sanding a wooden piece with sandpaper wrapped around wooden piece | +---/b4 ………… Action sound source (bursting and breaking) | +---/sticks Breaking disposable wooden chopsticks by hands | +---/aircap Crashing an air cap by hands | +---/b5 ………… Action sound source (clapping sound) | +---/clap1 Clapping (once) | +---/clap2 Clapping (once (no. 2)) | +---/claps1 Clapping (a plurality of clapping) | +---/claps2 Clapping (a plurality of clapping (no. 2)) | +---/clip1 Slamming a small clip | +---/clip2 Slamming a large clip | +---/cap1 Making cap A burst open | +---/cap2 Closing cap B | +---/snap Beating paper-made bellows (large and small) (used for beating a person's head in a TV laugh-in) | +---/cracker Setting off a cracker | +---/c1 ………… Characteristic sound source (small metal articles) | +---/bell1 Ringing a suspended small bell (single) by pulling the cord | +---/bell2 Ringing a suspended large bell (single) by pulling the cord | +---/bells1 Ringing suspended small bells (multiple) by pulling the cord | +---/bells2 Ringing suspended large bells (multiple) by pulling the cord | +---/bells3 Ringing suspended small and large bells (multiple) by pulling the cord | +---/coin1 Dropping a coin (single) on a wooden board (large) | +---/coin2 Dropping a coin (single) on a plywood board | +---/coin3 Dropping a coin (single) on a metal board | +---/coins1 Dropping coins (multiple) on a wooden board (large) | +---/coins2 Dropping coins (multiple) on a plywood board | +---/coins3 Dropping coins (multiple) on a metal board | +---/coins4 Shaking coins (multiple) in a container | +---/bells4 Ringing a handheld bell | +---/bells5 Ringing a bicycle bell | +---/tabbouri Ringing a tambourine | +---/c2 ………… Characteristic sound source (paper) | +---/book1 Dropping books and magazines A, B, C, D and E on paper | +---/book2 Dropping books and magazines A, B, C, D and E on paper (no. 2) | +---/tear Tearing copy paper | +---/crumple Crushing copy paper by hand | +---/c3 ………… Characteristic sound source (musical instruments) | +---/castanet Clicking castanets | +---/maracas Shaking maracas | +---/horn Blowing a bugle | +---/drum Beating a drum | +---/cymbals Striking cymbals | +---/string Twanging of a stringed musical instrument | +---/whistle1 Blowing whistle A | +---/whistle2 Blowing whistle B | +---/whistle3 Blowing whistle C | +---/ring Ringing a bell by shaking | +---/kara Shaking a baby rattle | +---/c4 ………… Characteristic sound source (electronic sound) | +---/phone1 Ringing of a home telephone | +---/phone2 Beep of a cellular phone | +---/phone3 Beep of a microcellular phone | +---/clock2 Ringing of an electronic alarm clock | +---/pipong Sound of an electronic sound toy A (pipong) | +---/buzzer Sound of an electronic sound toy B (beep) | +---/phone4 Beep of a cellular phone (no. 2) | +---/toy2 Sound of an electronic sound toy C (Gyuun) | +---/c5 ………… Characteristic sound source (mechanical) +---/stapler Stapling copy papers with a stapler +---/punch Punching copy paper with a punch +---/padlock Opening and closing of padlocks A and B +---/toy Sound by releasing spring +---/clock1 Ringing of a bell-alarm clock +---/doorlock Opening and closing of door locks A, B, and C +---/coffmill Grinding coffee beans with an electric grinder +---/shaver Sound of electric shavers A and B +---/dryer Sound of hair dryers A and B +---/mechbell Ringing of mechanical bell of a bicycle +---/reconst ………… Reconstructed signal +--- ir030.dat Impulse response (Reverberation room A [Reverberation time : 0.30sec.]) +--- ir078.dat Impulse response (Meeting room [Reverberation time : 0.78sec.]) +--- ir130.dat Impulse response (Reverberation room B [Reverberation time : 1.30sec.]) | +---/org ………… Dry source | +--- aircap.raw Crashing an air cap by hands | : : (100 signals) | +--- wood3.raw Beating handheld wooden board (large) A and wooden board (large) B against each other | +---/ir030 ………… Reverberation room A [Reverberation time : 0.30sec.] | +--- aircap.raw Crashing an air cap by hands | : : (100 signals) | +--- wood3.raw Beating handheld wooden board (large) A and wooden board (large) B against each other | +---/ir078 ………… Meeting room [Reverberation time : 0.78sec.] | +--- aircap.raw Crashing an air cap by hands | : : (100 signals) | +--- wood3.raw Beating handheld wooden board (large) A and wooden board (large) B against each other | +---/ir130 ………… Reverberation room B [Reverberation time : 1.30sec.] +--- aircap.raw Crashing an air cap by hands : : (100 signals) +--- wood3.raw Beating handheld wooden board (large) A and wooden board (large) B against each other
[non-speech top]
[sounds]
[sample]
[research 1]
[research 2]
[system]