Measurement of quality of service

[0001] This application is a continuation-in-part of U.S. application Ser. No. 09/870,366 filed on May 30, 2001 now U.S. Pat. No. 7,020,093, the contents of which are herein incorporated by reference.

[0002] This disclosure relates to measurement of quality of service.

[0003] In a content delivery system, a content-server transmits a media stream to a client over a communication channel. The content-server often transmits media streams at times during which the client system is unlikely to be in use. These media streams are saved in a mass-storage medium for later retrieval and viewing by an audience.

[0004] During transmission, transmission errors are introduced. These errors affect the viewability of the media stream. The extent of these errors is reflected in the “Quality of Service”, or QOS, for that transmission. In many cases, a content-delivery system measures QOS during transmission of the media stream. If the measured QOS indicates excessive transmission errors, then the content-server re-transmits the media stream.

[0005] A conventional content delivery service measures its QOS by collecting network statistics and inferring, on the basis of those network statistics, how good the media stream would appear to the viewing audience.

[0008] As shown in the example of FIG. 1, a content delivery system 10 for the delivery of a media stream 12 from a content server 14 to a client 16 first reduces bandwidth needed for transmission by passing the media stream 12 through an encoder 18 executing on the content server 14. The encoder 18 transforms the media stream 12 into a compressed form, herein referred to as the “encoded media stream 20,” suitable for transmission. The encoded media stream 20 then traverses a communication channel 22 until it reaches the client 16, whereupon it becomes a received encoded media stream 21. As shown in FIG. 1, the communication channel 22 is a wireless link between the client 16 and the content server 14. However, the communication channel 22 can also include a portion of a cable distribution network or a computer network. Examples of computer networks include WANs, LANs, private networks, and public networks such as the internet. This received encoded media stream 21 is stored on a storage device 27 for later viewing. In response to a request to view the media stream 26, or in response to a request for a QOS measurement, a decoder 24 executing on the client 16 retrieves the received encoded media stream 21 from the storage device 27 and transforms it into a decoded media stream 26.

[0009] In the content delivery system 10 of FIG. 1, there are at least two mechanisms that can impair the quality of the media stream. First, the encoder 18 and decoder 24 can introduce errors. For example, many encoding processes discard high-frequency components of an image in an effort to compress the media stream 12. As a result, the decoded media stream 26 may not be a replica of the original media stream 12. Second, errors can be introduced within the communication channel 22 itself. The nature of these errors depends on the type of communication channel. For example, in the case of satellite transmission, ionospheric conditions can degrade the received signal quality. Other mechanisms for introducing errors into the communication channel 22 include multipath reflection and dispersion.

[0010] These two impairment mechanisms, hereafter referred to as encoding error and transmission error, combine to affect the audience's subjective experience in viewing the media. However, the audience's subjective experience also depends on one other factor thus far not considered: the content of the media stream 12 itself.

[0011] In many cases, the extent to which a particular error affects an audience's enjoyment of a decoded media stream 26 depends on the content of the original media-stream 12. For example, a media stream 12 rich in detail will suffer considerably from loss of sharpness that results from discarding too many high frequency components. In contrast, the same loss of sharpness in a media stream 12 poor in detail, such as one having extensive night-time scenes, will most likely go unnoticed.

[0012] Referring to FIG. 2, a system 28 for measurement of QOS includes a content server 30 in data communication with a client 32. Communication between the client 32 and the content server 30 can be over a satellite link, as shown in the figure. Alternatively, communication can be established over a cable network, a global computer network such as the internet, or any other data communication network.

[0013] An encoder 38 applies an encoding or compression algorithm to the original media stream 39, thereby generating an encoded media stream 40. This encoded media stream 40 is them provided to the content server 30 for transmission to the client 32. In most cases, particularly in the delivery of video programming, encoding is carried out in advance and the encoded media stream 40 is transmitted to the client 32 at an off-peak time, such as in the middle of the night, when the client 32 is unlikely to be in use by an audience. In such cases, the encoded media stream 40 is stored on a mass-storage system (not shown) associated with the content server 30 to await transmission.

[0014] A variety of encoding processes are available. In many cases, these encoding processes are lossy. For example, certain encoding processes will discard high-frequency components of an image under the assumption that, when the image is later decoded, the absence of those high-frequency components will not be apparent to the viewing audience. Whether this is indeed the case will depend in to part on the features of the image.

[0015] In addition to being transmitted to the client 32, the encoded media stream 40 at the output of the encoder 38 is also provided to the input of a first decoder 42. The first decoder 42 generates first decoder output 43 by recovering the original media stream to the extent that the possibly lossy encoding performed by the encoder 38 makes it possible to do so.

[0016] The first decoder output 43 is then provided to a first feature extractor 44. The first feature extractor 44 generates first feature data 49 by implementing known feature extraction algorithms for extracting temporal or spatial features of the encoded media stream 40. Known feature extraction methods include the Sarnoff JND (“Just Noticeable Difference”) method and the methods disclosed in ANSI T1.801.03-1996 (“American National Standard for Telecommunications—Digital Transport of One Way Video Signals—Parameters for Objective Performance Specification”) specification.

[0017] The original media stream 39 is also passed through a second feature extractor 46, that performs feature extraction like the first feature extractor 44, that generates second feature data 47. The first and second feature data 49, 47 are then compared by a first analyzer 48. This comparison results in the calculation of an encoding metric. As noted above, a media stream can be degraded through transmission errors and through encoding errors. The encoding metric provides a measure of how badly the media stream is degraded as a result of encoding errors alone.

[0018] An analyzer 48 compares features of two media streams. The output of the analyzer 48 is typically a dimensionless quantity that represents a normalized measure of how different the two media streams would appear to a viewer. In some practices of the invention, the analyzer 48 is configured to re-encode some or all of the original media stream 39 if the encoder metric indicates that the first and second feature data 49, 47 are too different from each other.

[0019] The first feature data 49 is also provided to the content server 30. The content server 30 transmits both the encoded media stream 40 and the first feature data 49 to the client 32 by way of the first antenna 34. The encoded media stream 40 and the first feature data 49 can be transmitted concurrently or at separate times.

[0020] As it propagates between the first antenna 34 and a second antenna 35 associated with the client 32, the encoded media stream 40 is subjected to the various difficulties that are commonly encountered on a communication channel. These difficulties are manifested as jitter, packet loss, and packet latency in the encoded media stream 40 received by the client 32. In one embodiment, statistics on these and other measures of transmission error are collected by a network performance monitor 52 located at the client 32 and made available to a second analyzer 60 located at the client 32.

[0021] The media stream received by the client 32, referred to herein as the “received encoded media stream 53, is then stored on a mass-storage device 57. In response to a request to view the media stream, a copy of the received encoded media stream 53 is provided to a second decoder 54. The output of the second decoder 54, referred to herein as the “second decoder output 56,” is provided to a display 55 for viewing by an audience.

[0022] In some cases, transmission error significantly impairs the quality of the received encoded media stream 53. In cases in which the original media stream 39 is transmitted to the client 32 in advance of when it is viewed, there is an opportunity to correct this by re-transmitting some or all of the original media stream 39. For this opportunity to be taken advantage of, the client 32 must determine whether the received encoded media stream 53 has been significantly impaired.

[0023] In contrast to conventional systems, the client 32 does not simply examine network statistics during transmission to assess the impairment of the media stream. As noted above, whether the media stream is so impaired as to degrade the viewer's experience depends, to a great extent, on the content of the media stream. The client 32 instead compares first feature data 49 from the first feature extractor 44 with corresponding third feature data 59 extracted from the second decoder output 56 by a third feature extractor 58 that performs feature extraction like the first and second feature extractors 44, 46. The first and third feature data 49, 59 are then provided to the second analyzer 60 for comparison with each other.

[0024] Unlike the second decoder output 56 provided to the third feature extractor 58, the input to the first feature extractor 44 was never subjected to the vagaries of transmission. Hence, any difference between the first and third feature data 49, 59 is attributable to transmission errors alone. This difference is determined by the second analyzer 60 on the basis of a comparison between the first and third feature data 49, 59. On the basis of this difference, and optionally on the basis of network statistics 62 provided by the network monitor 52, the second analyzer 60 calculates a transmission metric 64 indicative of the extent to which the subjective perception of a viewing audience would be degraded by the transmission error alone. This transmission metric 64 can be sent back to the content server 30, either using the same channel that was used to transmit the encoded media stream 40, or through an alternate data communication channel, for example through a telephone line.

[0025] Upon receiving the transmission metric 64, the content server determines whether the QOS measured during transmission of the encoded media stream 40 is below a threshold. If the transmission metric 64 indicates that QOS during transmission was poor, the content server 30 re-transmits the encoded media stream 40. Otherwise, the content server 30 need do nothing further.

[0026] In an alternative embodiment, a decision to request retransmission on the media stream is made at the client 32 on the basis of the output of the second analyzer 60. In this case, the client 32 sends a signal back to the content server 30 to request re-transmission of the media stream. This further shifts the computational burden to the client 32 from the content server 30.

[0027] The client 32 thus provides an estimate of how a viewing audience is likely to perceive a second decoder output 56 derived from the received encoded media stream 53. If the received encoded media stream 53 proves to be excessively impaired, the client 32 requests re-transmission of the encoded media stream 40. In effect, the client 32 previews second decoder output 56 to determine whether it is of adequate quality to present to a viewing audience.

[0028] Instructions for carrying out the method described herein are typically stored on a machine-readable medium for execution by a processing element such as that found in digital computers, PDA's, and other devices that employ processing elements such as microprocessors and micro-controllers. Such devices can include electronic devices, optical devices, or combinations thereof.