H. H. Lee, C.K. Un, A study of On-Off Characteristics of Conversational Speech, IEEE Trans. on Communications, COM-34, num. 6, p. 630 (1986).  Home/Search   Document Not in Database   Summary   Related Articles   Check

....In telecommunication systems, service time distributions with coefficient of variation higher than one are often found. Squared coefficients of variation as high as 4.8 are common in fix telephony [5] When voice is multiplexed at talk spurt level the squared coefficient of variation reaches 2. 5 [6]. In Private Mobile Radio (PMR) systems which integrate dispatch calls (with a typical duration of 20 seconds) and interconnection to public telephone network (120 seconds) the coefficient of variation of the service time can be any, depending on the proportions of the call mixture. The same can ....

H. H. Lee, C.K. Un, A study of On-Off Characteristics of Conversational Speech, IEEE Trans. on Communications, COM-34, num. 6, p. 630 (1986).

....=6.934 s 2 =0.567 p=0.545 0 2 4 6 8 10 12 14 16 18 20 s 0 1 2 3 4 5 10 4 Exponential Lognormal 2 Density Figure 6. Transmission holding time distribution and fitting distributions. If we compare these results with those of Digital Speech Interpolation (DSI) the differences are clear. We see in [12] that the talkspurt duration can be modelled by two weighted geometric distributions (discrete version of an H 2 ) or by an exponential distribution when using a fill in time greater than 200 ms. Although transmission trunking and DSI make use of the same principle, increase the channel capacity ....

H.H. Lee, C.K. Un, "A Study of On-Off Characteristics of Conversational Speech", IEEE Trans. Comm., COM-34, No. 6, pp. 630-637, June 1986.

....taken to evaluate PRMA is based on measurements of recorded speech with a mean of 1 second for the talkspurt duration and 1.35 s. for the silence gap. There exist a standard model for telephone transmission by speech interpolation in the ITU T recommendation P. 84 [7] based on measures taken by Lee [8] that we will use instead along this work. We prove that the difference between both models is not small. b)The performance target The consequence on the subjective effects of limiting the drop probability to 1 is not well known for speech interpolation [5] When voice is packetized and ....

H.H. Lee, C.K. Un, "A study of On-Off Characteristics of Conversational Speech", IEEE Transactions on Communications, COM-34, no. 6, pp. 630-637, June

....user silence in the 28 messages where only one partner is heard, the mean is of 5080 ms, not very far from the 5670 ms we have found, so this figure can be considered quite accurate. If we compare these results with those of Digital Speech Interpolation (DSI) the differences are clear. We see in [8] that the talkspurt duration can be modelled by two weighted geometric distributions (discrete version of an H 2 ) or by an exponential distribution when using a fill in time greater than 200 ms. Although transmission trunking and DSI make use of the same principle, increasing the channel capacity ....

H.H. Lee, C.K. Un, "A Study of On-Off Characteristics of Conversational Speech", IEE Trans. Comm., COM-34, June 1986, pp. 630-637.

....In telecommunication systems, service time distributions with coefficient of variation higher than one are often found. Squared coefficients of variation as high as 4.8 are common in fix telephony [5] When voice is multiplexed at talk spurt level the squared coefficient of variation reaches 2. 5 [6]. In Private Mobile Radio (PMR) systems which integrate dispatch calls (with a typical duration of 20 seconds) and interconnection to public telephone network (120 seconds) the coefficient of variation of the service time can be any, depending on the proportions of the call mixture. The same can ....

H. H. Lee, C.K. Un, A study of On-Off Characteristics of Conversational Speech, IEEE Trans. on Communications, COM-34, num. 6, p. 630 (1986).

....time distribution of some telephone systems can be of the heavy tailed type (decays as a negative power of the time) and the coefficient of variation extremely large. The exponential distribution is not heavy tailed (decays faster) while the lognormal distribution is. According to Lee and Un [9], the duration of the talkspurt measured in discrete time intervals when the human voice is processed at talkspurt level can be represented by a combination of two geometric distributions. This is the model recommended by the ITU T (Rec. P.84) for systems that assign a channel for the talkspurt ....

H.H. Lee, C.K. Un, "A study of On-Off Characteristics of Conversational Speech", IEEE Trans. on Communications, COM-34, num. 6, pp. 630-637, June 1986.

....of the request packets 3. The size of the retrieved objects 4. Time to parse the main HTML object 5. Number of inline objects in an HTML document 6. User think time in viewing a web page. 3 Previous Modeling Work In the last few years, a number of Internet traffic modeling papers have appeared [5,6,7,9,10,11], which have primarily focussed on the aggregate characteristics of the traffic, based on the protocol used. It has been recently shown[11] that aggregate WWW traffic exhibits self similarity, which has been inferred from the underlying distributions of the WWW document sizes, the effect of user ....

....This helps in estimating the number of acks sent out by the client. We have observed that both the main object as well as the inline object, have similar distributions for the number of tcp data segments transfered from server to the client. The mean number of segments for the main document is 8. 3 [9] with a median value of 4. The mean number of segments for the inline object is 7.27 [8] with a median of 3. The inlined objects tend to be slightly smaller as they are usually compressed image files used for the icons in a web page as compared to the uncompressed ASCII HTML file (main object) ....

[Article contains additional citation context not shown here]

H. H. Lee and C. K. UN, "A study of On-Off Characteristics of Conversational Speech",IEEE transactions on Communications, Vol. Com-34,No.6,June 1996.

.... change over time (such as in an automobile) Additional knowledge about speech signals can be added to the algorithm so that speech can be differentiated from transient background sounds [27] For example, speech must include breath pauses, and these occur with well known timing characteristics [25]. Such information could help distinguish a passing train from a short monologue. It is possible to dynamically adapt the segmentation algorithm based on the content of the recording rather than using fixed parameters. For example, in determining the segments for level 3 skimming it may be better ....

Lee, H.H. and Un, C.K. A Study of on-off Characteristics of Conversational Speech. IEEE Transactions on Communications COM-34, 6 (Jun. 1986), 630--637.

....protocols provide adequate performance for both voice and data transmission at the cost of occasional losses (but within strict delay limits) for the former and reasonable delays (but with no loss) for the latter traffic. Voice is statistically multiplexed using speech activity detectors (SAD) [3, 7], and a reservation is set up when the voice user is active (i.e. talkspurt state) and it is released when he she becomes inactive (i.e. silence state) during the transmission. Channel resources are used more efficiently This work was supported by the Bogazi ci University Research Fund under ....

H. H. Lee and C. K. Un, "A study of on-off characteristics of conversational speech", IEEE Transactions on Communications, vol. 34, June 1986, pp. 630-637.

No context found.

H.H. Lee, C.K. Un, "A study of On-Off Characteristics of Conversational Speech", IEEE Transactions on Communications COM-34, no. 6, pp. 630-637, June 1986

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC