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
  • G10L25/69—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
• • 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
  • G10L21/00—Speech or voice signal processing techniques to produce another audible or non-audible signal, e.g. visual or tactile, in order to modify its quality or its intelligibility
  • G10L21/02—Speech enhancement, e.g. noise reduction or echo cancellation
  • G10L21/0208—Noise filtering
  • G10L21/0216—Noise filtering characterised by the method used for estimating noise
  • G10L2021/02168—Noise filtering characterised by the method used for estimating noise the estimation exclusively taking place during speech pauses

Definitions

• the invention relates to instrumental methods for measuring the speech quality of recorded or transmitted voice signals. Doing so will measure of speech quality assumed that e.g. with the ITU-T standard P.862 ("Perceptual Evaluation of Speech Quality (PESQ), an Objective Method for end-to-end Speech Quality Assessment of Narrow-band Telephone Networks and Speech Codecs ", ITU-T, Geneva, 2001) become.
• P.862 Perceptual Evaluation of Speech Quality (PESQ), an Objective Method for end-to-end Speech Quality Assessment of Narrow-band Telephone Networks and Speech Codecs
• the perceived speech quality z. B. for telephone connections or radio transmissions is mainly caused by simultaneous speech disorders, i.e. disorders during the Speech activity, determined. But noises in speech breaks also go into that Quality assessment, especially with high quality speech reproduction quality.
• Speech quality determinations of speech signals are generally made using auditory ("subjective") investigations with test persons.
• the goal of Instrumental ("objective") procedures for determining speech quality are made up of Properties of the speech signal to be evaluated by means of suitable computing methods To determine characteristic values that determine the speech quality of the speech signal to be assessed describe without having to resort to judgments from test subjects.
• Known methods for instrumental determination of speech quality determine the speech quality based on a comparison between undisturbed reference speech signal (source speech signal) and the to be evaluated and possibly disturbed signal.
• sample connection systems at which a known reference speech signal (source speech signal) is fed in at the source and z. B. transmitted over a telephone connection and recorded at the sink. After The speech signal is recorded and a speech quality value is calculated.
• source speech signal source speech signal
• Instrumental procedures for determining language quality are usually limited to Evaluation of sections with language activity.
• the current ITU-T standard P.862 is also only limited to sections with the determination of the speech quality active language. Especially with high quality speech reproduction and noises that these methods (e.g. measuring methods subsequently only occur during pauses in speech) ITU-T Rec. P.862) unreliable quality values.
• the speech quality here too optimistic because the speech quality felt by a listener is based on the entire signal including possible noises in the speech pauses.
• Some instrumental procedures for determining speech quality such as B. the procedure according to ITU-T Rec. P.862, take into account the calculation of the speech quality values No sounds during the pauses in speech. The resulting measured values are especially with high playback quality in the case of voice activity but occurring noises during pauses in speech, unreliable. With the present method, the background noises are intended be taken into account in the pauses in the speech when determining the speech quality values.
• the solution to the task assumes that the background noise in the Speech breaks regarding their disturbing influence on the perceived speech quality be rated.
• intensity characteristics of the background noise are determined and with these values the speech quality measured values, which are determined by an instrumental Process, e.g. B. according to ITU-T Rec. P.862, were corrected.
• the speech quality value using the source speech signal and the evaluating disturbed speech signal e.g. B. with the method according to ITU-T Rec. P.962, calculated.
• These speech signals are also available to the subsequent correction process Input parameters available.
• the correction procedure described here is required still the calculated speech quality value, the z. B. with the method according to ITU-T Rec. P.862 was calculated.
• one or more intensity parameters are generated of the noise in pauses in speech.
• This can e.g. B. the average loudness according to ISO 532 of the background noise during speech pauses.
• Other intensity parameters e.g. Sharpness, impulsiveness, fluctuation strength
• sharpness, impulsiveness, fluctuation strength can be included in the correction value. It it is assumed that increasing intensity values of one also increasing disturbance caused by the noise in speech pauses and thus lead to a greater reduction in the perceived speech quality.
• the speech quality is rated too high by the described speech quality measurement methods.
• One or more background noise intensity values during speech pauses are used to correct the measured speech quality value. Assuming that the measured speech quality value on a scale from 1 (low quality) to 5 (very good quality) according to ITU-T Recommendation P.800 ("Methods for objective and subjective assessment of quality", ITU-T, Geneva 1996 ), all values above a certain speech quality threshold value (e.g. above 3.0 in the ITU-T Rec. P.862 procedure) are reduced if background noise occurs during speech pauses.
• This reduction depends on the intensity characteristics of the background noise, the proportion of the speech pauses in the overall signal PA and the speech quality value Y.
• the function a (N) represents a clear and increasing weighting function of the intensity characteristic value N.
• the corrected speech quality value is always less than or equal to the uncorrected value.
• the correction is small if the intensity of the noise is low (N is small), there are only a few pauses in speech ( PA small) or the speech quality value is close to the speech quality threshold ( Y-YS small). Corrections are made more strongly if there are strong pause noises with an otherwise high speech quality Y.
• the intensity parameters are weighted with the weighting functions a (N), b (M) and c (O) . Since an increasing disturbance with increasing values is assumed, clear and increasing weighting functions must also be used here.
• the exemplary embodiment presented here shows an example of a correction of the speech quality values determined using the method according to ITU-T Rec. P.862 "PESQ" (status 2001).
• This method provides a speech quality value by comparing an undisturbed source speech signal with the disturbed speech signal to be evaluated. These two speech signals are used to determine the mean loudness of the background noise in accordance with patent application DE 101 20 168.
• the value N in sone thus calculated is used to correct the quality value calculated using the method according to ITU-T Rec. P.862 "PESQ” (status 2001) used.
• the value of the threshold crossing YD is determined, ie by what proportion the measured speech quality Y exceeds the speech quality threshold YS .

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)
• Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
• Monitoring And Testing Of Exchanges (AREA)
• Noise Elimination (AREA)

Applications Claiming Priority (2)

Publications (3)

Family

ID=7697364

Family Applications (1)

Country Status (3)

Cited By (1)

Family Cites Families (7)

• 2001
  • 2001-08-29 DE DE10142846A patent/DE10142846A1/de not_active Withdrawn
• 2002
  • 2002-06-10 DE DE50205328T patent/DE50205328D1/de not_active Expired - Lifetime
  • 2002-06-10 AT AT02012790T patent/ATE313846T1/de not_active IP Right Cessation
  • 2002-06-10 EP EP02012790A patent/EP1288914B1/fr not_active Expired - Lifetime

Also Published As

Similar Documents

Legal Events