: 3.1 Sound control Emergency Stop device
by Kiyoshi FUKAYA
The development of automation has brought about many changes in factories. One of them is underpopulation of workers, i.e. work range has broadened and no colleague is around the worker in automated line. When such a worker happen to meet an accident, even though he cries for help, he can not expect help from colleagues. In these circumstances, a device which recognizes voice for help and stops the machine will be great help. The possibility of application of speech analysis technology to emergency stop device was investigated.
The system consists of two sub-system : voice input system and voice recognition system. The system must have the function of distinguishing voice from background noise. In this research, the input system bears this function.
Voice input system consists of microphone and radio wave transmitter and receiver. In order to give worker the freedom of movement, use of radio device is necessary.
In order to lessen the influence of noise, the bone transmitting microphone was used. It detects the vibration which arise in vocal cords and is transmitted through head bone, and therefore it has low sensitivity for sound. The measurement of sensitivity showed sufficient faculty.
Two voice recognition devices of speaker dependent type were tested. Recorded voices were registered and were recognized in various playback condition ; normal, low/high tone, low/high speed, small/large voice, etc. They recognized poorly voices which were not played back but pronounced each time and voices which were played back in high speed. The shift of voice to high pitch resembles the voice in time of emergency. It seemed that they can not cope with the fluctuation of human voice.
In order to distinguish stop command, simple algorithm was adopted : large voice which lasts for a certain duration considered as command for emergency stop. Simple device which fulfills the algorithm was made and tested. The test showed that the total system has the possibility of practical use.
The device developed do not have fail safe function, which is indispensable to "safety device". Therefore this is not safety device but safety assist device. To give them the fail safe function is future problem.
3.2 Selection of Auditory Warning Alarm According to Different Unsafe Condition
by Yoshiyuki EGAWA
For evaluating the characteristics of auditory warning which were safety measurements using acoustics, this: report investigated them from the two points of view as follows.
I. Warning alarm should have the characteristics not to be masked by ground noise in the factory.
II. Warning alarm should be made up of wave forms which are suitable for utility purposes of warning usage.
About I theme, the data which we had measured about fifty kinds of ground noises in many factorys were classified as eight patterns model-noise by cluster analysis. According to form in 1/3 Oct. analysis, their models were named "Pink-noise, 2 tsu-yama (koiki), Chuiki-yama, Teiiki-ochi, Owan, Chuiki-ochi, Koiki-yama, and Teiiki-yama."
Next, spectrum analysis of warning alarm being made use of various occasions was performed. As the result analysis, four types of alarms were found in auditory warning.
From a frequency band viewpoint, it was found that warning alarm had two types which were Wide band type and Narrow band type.
Further, there were two types, which were named Noise spect.-type and Linear Spect.-type from a spectrum form.
Experiments on discrimination four types alarm from eight patterns model-noise were conducted. As the result of experiments on discrimination, Wide band type alarm was more available than Narrow band type in all model-noises and Noise spect.-type alarm was more discriminative than Linear spect.-type in Chuiki-yama model noise.
Regarding as shift type alarm (like a siren) being made use of emergency car (i.e. fire engine) , the siren was most discriminative in the case when the parameter indicated 1000 ˜ 2000 Hz/2 sec.
About II theme, we arranged auditory warning alarm according to the wave form and purpose of warning usage.
This result was illustrated in Table 3.2-5.