Is it possible to whistle underwater

This got us wondering about other amazing products that have achieved this feat of engineering. This baffles audiences every time they see it in a movie. Surely flames and water do not mix? Yet they do. Underwater flares emit high-pressure jets of compressed air at the tips which produce a protective bubble, enclosing the flame and shielding it from the water.

PvtStash Contributor. Messages Reaction score 8 Location Toledo, Ohio of dives - AbyssalPlains said:. Click to expand CuracaoJ Contributor. Messages Reaction score 67 Location Curacao of dives. I sort of agree with Devon and Abyssal, but then again, I sort of don't I agree that a noise maker should not ever be a replacement for good buddy skills and practices.

I also agree that they should never be allowed to be used as a crutch. One should never think "oh my buddy isn't looking but I'm going to swim 10 meters away and go look at xyzthingy, she or I can rattle a noise maker if we need each other. I also keep a spare in my bag for other buddies. Good buddy skills should be the norm, but lets face it every once in a while things happen, and people get distracted. Occasionally they are useful to get the attention of other divers besides my buddies.

A few dives back I had to grab another divers attention in a wreck that was about to wrap his tank into a wire going across a room. He was about 5 meters away, and I was able to get his attention and have him notice the wire. I look at it sort of like my knife. I don't ever want to be diving in a mess on monofillament, but if it happens I want to have a knife with me.

The other thing is it can be useful to say "hey you are about to get eaten by a giant swimming lobsterman from behind! My two BAR Jeff. Thanks for the info and advice. Mopar Contributor. Thus, a low BER performance in the ocean experiments and covertness, i.

In this paper, we propose a biomimetic covert communication scheme that modulates the information bits into various whistle patterns to increase the DoM with the link information among consecutive transmitted whistles, and detects the distorted whistles—via the link information—with the UWA channel using a machine learning based detector.

The proposed method divides a large number of dolphin whistles into groups based on the similarity of the patterns. Each group is used as a symbol and mapped to information bits. To convey information bits and maximize the DoM, the randomly selected whistle in a chosen group is transmitted, and different whistles are sequentially transmitted.

When a number of transmitted whistles pass through the UWA channel and background noise is added, the conventional machine learning based detectors suffer from detecting many distorted whistles. For a large the DoM, we directly utilize many real dolphin whistles for the modulation. For a small BER, we optimally classify the real dolphin whistles with large distances, and develop a trellis structured transmission algorithm using the information link matrix, without sacrificing the DoM.

For increasing the detection performance of the nonlinear characteristics of many transmitted whistles, we develop a DAG-net based machine learning detector.

The performance of the proposed algorithm has been proved through computer simulations and practical ocean experiments. The paper is organized as follows.

Section 2 describes the characteristics of the dolphin whistles and the whistle classification by groups. Section 3 proposes the modulation method that allocates bit information into whistles using the link information.

In Section 5 , the learning process of the proposed method is shown. Section 6 demonstrates the BER performance by using computer simulations and practical ocean experiments.

Section 7 concludes the paper. Dolphins communicate with each other using whistles. The general dolphin whistles have a time duration that varies from several hundred milliseconds to two seconds, and a frequency bandwidth that varies from several hundred Hz to tens of kHz [ 16 , 17 , 18 , 19 , 20 ]. The variation of frequencies over the time duration is referred to as the frequency contour or whistle pattern [ 16 , 17 , 18 , 19 , 20 ].

In Figure 1 and Figure 2 , the whistle spectrograms of the false killer whales and white sided dolphins, respectively, are displayed. The frequency components of the dolphin whistles vary in time. In Figure 1 and Figure 2 , many dolphins generate many different or similar whistle patterns. The spectrograms of the false killer whale whistles. The spectrograms of the white sided dolphin whistles.

In Figure 1 and Figure 2 , the similar whistle patterns are marked by rectangular and circular boxes. In Figure 1 , the rectangular and circular boxes contain up-chirps with a different frequency variation and a large variation, respectively. In Figure 2 , the rectangular and the circle boxes are marked for flat-downward scoops and down chirps, respectively.

In practice, more whistle patterns than in the above examples can be found. If the whistles with similar patterns are classified as the same group and binary bits are allocated to the whistle group, we can transmit binary information with the same dolphin whistle patterns.

Thus, the proposed method transmits one of the randomly selected whistle patterns in the group, which preserves a larger DoM and greater covertness than the conventional biomimetic UWA communication methods. When all whistles are classified as groups, the distance between groups needs to be kept as far as possible to attain the low BER. Thus, we classify the whistles to maximize the distance based on the whistle features.

For whistle classification with the maximum distance, firstly, we set a whistle feature vector V whose elements present the dominant features of the whistle, e.

V is set as below,. If V consists of the j elements, V has a vector space of R j and the classification with the maximum distance is performed in R j space. For the classification, we change the classification problem of maximizing distances among different whistle groups, to a new problem of minimizing the distances in the same groups. Since the k-means clustering algorithm is known as a good classification method for minimum variance [ 19 , 20 ], the k-means clustering algorithm is utilized for classifying the whistles as groups.

Then, the k-means algorithm for classifying the whistles is written as [ 21 , 22 ],. Equation 2 minimizes the variance of the vectors V n which belongs to the same group, i. Then, K opt groups are obtained, and the whistles in the same group have similar patterns.

When a large number of information bits are inputted and all whistles are transmitted, the large DoM is attained.

At the receiver, the conventional maximum likelihood ML based detector can be used to estimate the group index k from the received whistles. When the conventional ML detector detects the received whistles, the ML detector suffers from two problems: for the first problem, when the number of real whistles is large, a low BER is not obtained. The maximum distance by the k-means algorithm between groups may not be large enough to overcome the background noise and the UWA channel distortions.

If some whistles in the group are picked to keep the larger distance and the error correction schemes are used, the BER performance can increase.

However, the DoM and the data rate decreases. For the second, when the received whistles are distorted by the UWA channel, the distorted whistles cannot be compensated by the equalizer. Even though the channel is estimated by the pilots, the frequency bandwidth and time duration of the whistles are larger than the coherent frequency and the coherent time, respectively. In addition, the repeatedly transmitted pilots reduce the DoM.

The small distance problem may be resolved by the modern machine learning detection scheme, whose detection capability is better than the ML. When the whistle patterns are represented as a 2-D image, e.

However, the distorted whistle problem of the UWA channels may not be overcome by the conventional machine learning methods. This is because the detection accuracy of the conventional machine learning methods may not be enough to satisfy the common BER requirements of the communications. Therefore, we propose a biomimetic communication scheme: the transmitter modulates the whistles with the larger distances and DoM based on the link information among adjacent whistles, and the receiver demodulates and detects the distorted whistles using a DAG-net based LSTM with additional link information among whistles.

The proposed modulation and demodulation method are described in the following sections. The FEC provides a connection rule to concatenated additional symbols and corrects the erroneous bits, but requires the additional bits that reduce the data rate [ 23 , 24 ].

In cases of small bandwidth UWA communication systems, the data rate is one of the important parameters. Thus, the proposed biomimetic communication scheme is developed to obtain a low BER and large DoM, without sacrificing the data rate. The proposed method reclassifies the real whistles into a large number of subgroups to have larger distances among subgroups and provide link information among subgroups to utilize all subgroups.

The detailed procedure and an example are described in this section. The possible number of the subgroup sets with the size of 2 M opt is very large, but in this paper, K tot subgroup sets that have large distances are chosen.

If we carefully select the subgroup sets whose distances are larger than those of the first classified K opt group sets, we attain a lower error rate performance. The i -th column index is mapped to input bits for the binary allocation, e. Then, the link information between two subgroup sets is established.

In order to avoid falling into a short loop by the subgroup loop indexing, the element value of a row of H is not allowed to include the same row index and the elements of the one subgroup set is not the same as that of other subgroup sets, and all indices occur evenly 2 M opt times.

This additional link information helps the receiver to decode the distorted received symbols and attain a better BER performance. A link matrix H is given as below,. When b 1 is input, the b 1 binary value is translated to a decimal number which is the index of the column of the first row.

The selected element of the first row indicates the subgroup index for b 1 , and a whistle in the indicated subgroup is randomly chosen and transmitted. Subsequently, the element selected by b 1 also provides the row number of H for the next input b 2. For b 2 , the same procedure is performed for the input b 1.

This procedure is executed to all elements of B. If the size of B is large, all elements of H will be picked and all whistles will be utilized for the transmission, which preserves the large DoM. No same subgroup set, e. Let the n -th transmitted whistle be W n , and the subgroup index of the whistle be X n. One whistle W 1 in the seventh subgroup X 7 is randomly selected and transmitted as the first symbol. One whistle W 4 of the first subgroup X 1 is randomly picked and the first row is chosen for the fifth input bit.

This sequential link result is depicted in Figure 3 b. When was Blow the Whistle - song - created? Why does a whistle whistle?

How do you play a whistle? How is Flo rider whistle rude? How do you blow the whistle on club penguin? What is the song with the techno whistle? What does blow the whistle mean? What rhymes with mistletoe? How do you blow club penguin whistle? Does the referee blow the whistle after an incomplete pass? What do umpiers blow to stop a game?

How do you whistle through your nose? At what age can a baby blow a whistle? Do you believe that you could undertake a whistle blow act? What happens if you blow on a whistle that was just in the garbage?

In the Legend of Zelda spirit trackers how do you blow the whistle? How do referees stop sports rage? How do you whistle with two fingers? Do you need to blow the whistle for a corner kick to be taken?