Classifications

• • G—PHYSICS
  • G10—MUSICAL INSTRUMENTS; ACOUSTICS
  • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
  • G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
  • G10L25/48—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 specially adapted for particular use
  • G10L25/51—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 specially adapted for particular use for comparison or discrimination
  • G10L25/60—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 specially adapted for particular use for comparison or discrimination for measuring the quality of voice signals
• • G—PHYSICS
  • G10—MUSICAL INSTRUMENTS; ACOUSTICS
  • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
  • G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
  • G10L25/48—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 specially adapted for particular use
  • G10L25/69—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 specially adapted for particular use for evaluating synthetic or decoded voice signals

Definitions

• This invention relates to methods and apparatus for providing quality measurements for voice equipment under test, and more particularly to methods and apparatus for providing real-time objective perceptual quality measurements of voice or audio signals received by such equipment.
• Voice quality evaluation is a difficult task for speech systems, especially those involving compression and coding, because common waveform and spectrum similarity criterion do not correlate particularly well with perceived quality of received voice signals.
• voice quality evaluation of telecommunication systems have been measured off-line using formal perceptual listening tests that are performed in a carefully controlled environment, using pre-prepared voice material. Although this practice is effective, it is both costly and time consuming. In addition, results obtained from such tests are dependent upon the individual test subjects and their environment. As a result, findings from such tests are not always repeatable or consistent.
• Known objective perceptual quality measurement systems require measurement of voice quality to be done off-line, i.e., from stored, received voice data. It would be desirable if such objective perceptual quality measurements could be made in real time or near real time in operational equipment.
• the present invention in one aspect, is a method for providing real-time perceptual quality measurements of an audio signal.
• a quality test signal including an audio test signal, is received by equipment under test.
• Playback of a pre-stored representation of the audio signal is coarsely synchronized to the received audio test signal, for example, utilizing a synchronizing pulse in a header of the quality test signal.
• the playback is then finely synchronized to the received audio signal, for example, by comparing data in a windowed portion of the received audio test signal and a windowed portion of the pre-stored representation of the audio test signal and by adjusting a windowed portion of the pre-stored representation of the audio test signal in accordance with results of the comparison.
• a window of the received audio test signal is then compared to a portion of the finely synchronized playback of the pre-stored representation of the audio test signal to output a quality measurement of the received audio test signal.
• the invention comprises an audio quality analyzer (AQA) to evaluate quality of a quality test signal received by equipment under test, where the quality test signal includes an audio test signal.
• the AQA is configured to coarsely synchronize playback of a pre-stored representation of the audio test signal to the received audio test signal, to finely synchronize playback of said pre-stored representation of the audio test signal to the received audio test signal and to compare a window of the received audio test signal to a portion of the finely synchronized playback of the pre-stored representation of the audio test signal to output a quality measurement of the received audio test signal.
• FIG. 1 is a block diagram of an embodiment of a voice quality analyzer in accordance with the invention.
• Figure 2 is a diagram of a quality test message frame.
• FIG. 3 is a diagram of another embodiment of a voice quality analyzer in accordance with the invention.
• Figure 4 is a block diagram of an embodiment of buffer providing sync windowing and selection windowing in accordance with the invention.
• Figure 5 is a drawing representing a rectangular window function shape.
• Figure 6 is a drawing representing a nonlinear emphasized window function shape.
• Figure 7 is a drawing representing a discontinuous rectangular window function.
• Figure 8 is a block diagram of a test configuration in accordance with the invention.
• Figure 9 is a flow chart of an embodiment of a test method in accordance with the invention.
• FIG. 1 is a block diagram of a voice quality analyzer (VQA) 10 that receives a voice signal output by voice equipment under test (VEUT) 12.
• VQA 10 comprises a quality evaluator 14 that generates a quality measurement of voice test signals received from VEUT 12.
• VQA 10 also comprises a header detector 16 which, in turn, comprises a dual tone multiple frequency (DTMF) detector 18 and a sequencer 20.
• DTMF detector 18 monitors signals received from VEUT 12 to detect and decode signaling tones present in the received signals. The decoded signals are used by sequencer 20 to control operation of a voice sentence generator 22.
• DTMF dual tone multiple frequency
• Pre-stored representations of voice test signals are stored in voice sentence generator 22. Such "sentences" may, but do not necessarily represent full sentences or words in any particular language, nor do they necessarily represent speech from any particular human. Rather, the representations are selected for facilitating the voice quality measurement performed by quality evaluator 14.
• sequencer 20 initiates playback of a particular pre-stored voice test signal representation from voice sentence generator 22, depending upon a particular voice test signal that is identified in the header.
• a fine synchronizer 24 is provided to achieve synchronization between the pre-stored representation of a voice test signal and the received voice test signal sufficient to perform an objective perceptual quality comparison utilizing quality evaluator 14.
• Voice quality measurement is performed by applying an objective perceptual quality measurement algorithm to compare the a portion of the synchronized, locally generated reference signal from fine synchronizer 24 to a windowed portion of the signal received from VEUT 12.
• one of the following algorithms is used: Perceptual Speech Quality Measure (PSQM), Measuring Normalizing Blocks (MNB), Perceptual Analysis Measurement System (PAMS), and Modified Bark Spectral Distortion (MBSD) measure.
• PSQM Perceptual Speech Quality Measure
• MNB Measuring Normalizing Blocks
• PAMS Perceptual Analysis Measurement System
• MBSD Modified Bark Spectral Distortion
• a plurality of different algorithms is available, and an algorithm selection is made manually.
• a plurality of different algorithms are available, and a selection is made dependent upon which pre-stored representation in voice sentence generator 22 is selected by sequencer 20.
• Quality test message 30 includes four sections 32, 34, 36, 38, of which three, 32, 34, and 36, comprise a header 40 that is transmitted utilizing DTMF signaling, and a fourth includes a voice test message 38.
• Unique word 32 is used to signal the start of a new quality test message 30.
• Unique word 32 is included to prevent false measurement start signals during periods of severe channel degradation, for example, periods of very noisy reception by VEUT 12 of signals from a cellular network.
• Sentence ID 34 includes an index number or identifier of voice test message 38, thereby permitting different test messages to be transmitted to VEUT 12 and identified by VQA 10.
• Sync pulse 36 is a short DTMF pulse that is used to signal the start of voice test signal 38.
• Sync pulse 36 is used by sequencer 20 to start voice sentence generator 22 playing the appropriate pre-stored voice test signal representation for comparison with that received by VEUT 12.
• header 40 is transmitted in another manner, for example, using another form of in-band signaling, or by using out-of-band signaling.
• means other than DTMF detector 18 are used to detect and respond to header 40.
• suitable in-band signaling include monotone signaling and telephony data protocol.
• An example of suitable out-of-band signaling is signaling on a separate paging channel.
• sequencer 20 includes a unique word detector 42, a sentence ID detector 44, and a coarse sync detector 46, which include the functions of DTMF detector 18 of Figure 1. Therefore, no separate DTMF detector 18 is shown in Figure 3.
• sentence ID detector 44 detects sentence ID 34 that is received after the unique word.
• sentence ID 34 is passed to voice sentence generator 22 so that it can output the proper pre-stored representation of a voice test signal corresponding to a voice test signal identified by sentence ID 34, and subsequently received data is passed to coarse sync detector 46.
• Coarse sync detector 46 detects sync pulse 36 which, in one embodiment, is coded as a short DTMF pulse. When a coarse sync signal from coarse sync detector 46 is received, voice sentence generator 22 begins playback of a pre-stored representation of a voice signal corresponding to the determined sentence ID 34.
• the coarse synchronization provided by sync pulse 36 is not sufficient to enable signal comparator 14 to compare a voice test signal 38 to a pre-stored representation of a voice signal in real time, i.e., so that the quality evaluations performed by signal comparator 14 occur during receipt of voice test signals 38 with little or no apparent delay as perceived by a user.
• coarse synchronization is not sufficient for analyzing voice test signals 38 using Perceptual Speech Quality Measure (PSQM), Measuring Normalizing Blocks (MNB), Perceptual Analysis Measurement System (PAMS), and Modified Bark Spectral Distortion (MBSD) measure algorithms. Therefore, a fine sync detector 24 is provided for more accurate synchronization.
• PSQM Perceptual Speech Quality Measure
• MNB Measuring Normalizing Blocks
• PAMS Perceptual Analysis Measurement System
• MBSD Modified Bark Spectral Distortion
• Fine sync detector 24 compares the output of voice sentence generator 22 with a window of voice data selected by sync windowing module 52. This comparison is performed, in one embodiment, in accordance with International Telecommunications Union (ITU) standard P.931 , "Multimedia Communications Delay, Synchronization and Frame Rate Measurement.” As a result of this comparison, outputs of fine sync detector 24 are produced to control a switch 54, which is closed when fine synchronization is achieved. Switch 54 prevents quality evaluations from being output before fine synchronization is achieved. In addition, data windows representing synchronized portions of a pre-stored representation of a voice test signal are output to a selection windowing module 56.
• ITU International Telecommunications Union
• Selection windowing module 56 selects a synchronized portion of the incoming voice test data 58 to compare to the synchronized portions of the pre-stored representation 60. The comparison is performed by perceptual comparator 14, and a quality evaluation is produced. The quality evaluation is output when switch 54 is closed, as indicated above.
• Figure 4 is a drawing of a representation of the windowing operation of sync window module 52 and selection windowing module 56 in one embodiment of the invention.
• a sync window 62 is selected from buffer 48 by sync window module 52.
• the start of sync window 62 and a selection window 64 selected by selection windowing module 56 are aligned.
• Buffer 48 is a circular buffer accepting digitized voice input.
• sync window 62 is adjusted in accordance with quality measurements made by perceptual comparator 14, as indicated in Figure 3. Alignment of selection window 64 with sync window 62 is accomplished, in this embodiment, by fine sync detector 24, including by selection of windowed data output from voice sentence generator 22.
• selection windowing module 52 also applies a window function to at least one of the received voice data and the pre-stored representation of voice test signals for data weighting.
• a plurality of weighting functions are provided, including rectangular weighting, as represented in Figure 5, nonlinear emphasized weighting, an example of which is represented in Figure 6, and discontinuous rectangular weighting, an example of which is represented in Figure 7.
• the selection of the weighting function is preselected, through selection of a quality algorithm.
• the selection is also adaptively alterable, in accordance with a quality measurement from perceptual comparator 14 and as indicated in Figure 3.
• Discontinuous rectangular weighting is used, for example, when disturbances such as hand-offs in a cellular system interfere with reception of voice signal data.
• the algorithm used by perceptual comparator 14 excludes the disturbed periods from the quality evaluation. The occurrence and length of disturbed periods, in one embodiment, is reported separately from the quality measurement.
• VQA 10 is shown as a computer in Figure 8.
• VQA 10 is connected to an output port of VEUT 12, which, in one embodiment, is a cellular telephone 12 with a "hands-free" port.
• VEUT 12 receives quality test messages 30 from a message source 66, for example, via a network 68 such as a cellular wireless network.
• message source 66 is configured as an answering machine with recorded quality test messages 30 stored in voice mailboxes.
• the recorded quality test messages 30 in the voice mailboxes are identified with sentence IDs 34.
• Voice test signals 38 stored in message source 66 are identified with sentence IDs 34 that identify corresponding pre-stored representations of voice test messages in voice sentence generator 22 of VQA 10.
• VEUT 12 dials 100 message source 66 via network 68 and retrieves 102 a voice mail message therefrom.
• the retrieved voice mail message is a quality test message 30.
• VQA 10 then waits 104, 106 until unique word 32 is recognized.
• sentence ID 34 is obtained 108.
• VQA 10 then waits until sync pulse 36 is received 110, 112.
• sync pulse 36 is received, a local copy of voice test signal 38 is retrieved 1 14, for example, from voice sentence generator 22. Fine synchronization 1 16 of the local copy of voice test signal 38 is then performed, and a voice quality measure is computed 1 18 until it is determined 120 that voice test signal 38 has ended.
• voice test signal 38 When voice test signal 38 has ended, the computed quality is displayed 122, and the end of the test is reached 124. In other embodiments, quality tests may be repeated manually or automatically.
• the invention described herein provide real-time perceptual quality measurement of voice signals.
• the invention is particularly suitable for performing such measurements utilizing algorithms that have previously not been known to be suitable for real-time measurements of signals.
• the invention is also particularly suited for providing real time perceptual quality measurements when a highly compressed voice signal is transmitted.
• embodiments described herein are applicable to quality measurements of voice signals, it will be recognized that the invention is also suitable for quality measurements of non- voice audio test signals as well.
• voice quality analyzer 10 is thus, more generally, an audio quality analyzer (AQA)
• voice test signal 38 is an audio test signal
• voice sentence generator 22 is an audio waveform generator (such as a digitized waveform generator)
• the pre-stored representations of voice test signals in the audio waveform generator are pre-stored representations of audio test signals.

Landscapes

• Engineering & Computer Science (AREA)
• Computational Linguistics (AREA)
• Signal Processing (AREA)
• Health & Medical Sciences (AREA)
• Audiology, Speech & Language Pathology (AREA)
• Human Computer Interaction (AREA)
• Physics & Mathematics (AREA)
• Acoustics & Sound (AREA)
• Multimedia (AREA)
• Quality & Reliability (AREA)
• Monitoring And Testing Of Transmission In General (AREA)
• Transmission Systems Not Characterized By The Medium Used For Transmission (AREA)
• Testing, Inspecting, Measuring Of Stereoscopic Televisions And Televisions (AREA)

Priority Applications (5)

Applications Claiming Priority (1)

Publications (1)

Family

ID=22272829

Family Applications (1)

Country Status (4)

Cited By (6)

Families Citing this family (2)

Citations (3)

Family Cites Families (3)

• 1999
  • 1999-05-25 AU AU40970/99A patent/AU4097099A/en not_active Abandoned
  • 1999-05-25 KR KR1020017000881A patent/KR100623214B1/ko not_active IP Right Cessation
  • 1999-05-25 JP JP2000620620A patent/JP4500458B2/ja not_active Expired - Fee Related
  • 1999-05-25 WO PCT/US1999/011473 patent/WO2000072306A1/en active IP Right Grant

Patent Citations (3)

Non-Patent Citations (2)

Also Published As

Similar Documents

Legal Events