TWELP 600 bps Vocoder

Details

(version 5.0) achieves outstanding speech quality and intelligibility, unmatched by any other vocoder at this bitrate.

For Digital HF Radio and other markets.

TWELP Technology Features. The vocoder is based on newest technology of speech coding called "Tri-Wave Excited Linear Prediction" (TWELP) that was developed by experts of DSPINI. 

TWELP technology is a new class of vocoders that differs from any other LPC-based vocoders by:

  • advance reliable method of pitch estimation
  • pitch-synchronous analysis
  • advance tri-wave model of excitation
  • newest quantization schemes
  • pitch-synchronous synthesis

Thanks to these unique features, TWELP technology provides significantly better speech quality than other well-known technologies—including AMBE+2, MELPe, ACELP, and others—at equivalent bit rates ranging from 300 bps to 4800 bps and beyond.

Additionally, unlike other low-bitrate vocoders (such as MELPe, for example), TWELP delivers much higher quality for non-speech signals, including sirens, background music, and similar audio.

Speech Quality. This is a comparison with the MELPe vocoder, which operates at 600 bps.
The TWELP 600 bps and MELPe 600 bps vocoders were tested using the ITU-T P.50 speech base in 20 different languages.

Note: 
We have updated the speech database by minimizing inter-speech pauses to eliminate their impact on the evaluation results. Therefore, the numbers obtained from the quality measurements using this updated speech database differ from those previously obtained with the original speech database, where speech pauses were not removed.

The ITU-T P.862 tool was used to evaluate speech quality in terms of PESQ scores:
 
The diagram demonstrates a huge difference in speech quality between the TWELP 600 and the MELPe 600 vocoders. Exact numbers are shown in the table below.
LanguageTWELP 600 bpsMELPe 600 bps
American 2.506 2.211
Arabic 2.522 2.168
British 2.540 2.270
Chinese 2.506 2.055
Danish 2.615 2.174
Dutch 2.487 2.145
Finnich 2.499 2.243
French 2.606 2.258
German 2.616 2.227
Greek 2.571 2.164
Hindi 2.642 2.358
Hungarian 2.583 2.293
Italian 2.770 2.417
Japanese 2.659 2.308
Norwegian 2.562 2.197
Polish 2.621 2.274
Portuguese 2.694 2.370
Russian 2.506 2.119
Spanish 2.627 2.322
Swedish 2.688 2.437
Average2.5912.251

A difference is on average 0.340 PESQ
 Speech Intelligibility. Here is a comparison with the MELPe vocoder, which operates at 600 bps. The TWELP 600 bps vocoder and the MELPe 600 bps vocoder were tested using the ITU-T P.50 speech database, covering 20 different languages.
STOI (Short-Time Objective Intelligibility) and ESTOI (Extended Short-Time Objective Intelligibility) metrics were used to assess speech intelligibility: 
 
The diagram illustrates a significant difference in speech intelligibility between the TWELP 600 and MELPe 600 vocoders in the STOI metric. Exact values are provided in the table below:
LanguageTWELP 600 bpsMELPe 600 bps
American 84.53 79.24
Arabic 83.07 78.50
British 82.02 75.67
Chinese 83.40 77.40
Danish 85.13 79.12
Dutch 83.37 77.04
Finnich 79.64 74.76
French 84.48 78.79
German 84.37 79.00
Greek 83.48 77.89
Hindi 83.97 78.35
Hungarian 84.27 78.14
Italian 83.15 78.03
Japanese 84.14 79.30
Norwegian 84.74 79.24
Polish 84.24 78.37
Portuguese 84.01 78.04
Russian 82.71 75.74
Spanish 82.46 77.82
Swedish 82.31 76.89
Average83.4877.90

A difference is on average 5.58 %
Considering that a low-bitrate vocoder is a nonlinear device that significantly distorts the spectrum of the original speech signal, the ESTOI metric provides more accurate assessments of speech intelligibility after vocoding:
 
The diagram shows a significant difference in speech intelligibility between the TWELP 600 and MELPe 600 vocoders in the ESTOI metric. Exact numbers are shown in the table below.
LanguageTWELP 600 bpsMELPe 600 bps
American 74.48 69.76
Arabic 74.28 70.04
British 72.80 67.54
Chinese 75.65 69.85
Danish 76.22 69.81
Dutch 74.36 69.58
Finnich 70.44 66.23
French 75.68 70.28
German 74.59 68.55
Greek 76.04 70.59
Hindi 72.77 67.87
Hungarian 73.59 67.46
Italian 73.28 68.60
Japanese 76.09 71.22
Norwegian 76.13 70.59
Polish 75.53 70.79
Portuguese 75.07 69.53
Russian 73.07 66.88
Spanish 74.07 70.33
Swedish 71.91 67.17
Average74.3069.13

A difference is on average 5.17 %
You can see that speech intelligibility testing confirms the significant superiority of the TWELP 600 bps vocoder over the MELPe 600 bps vocoder.
 
You can download the P.862 and STOI/ESTOI utilities, along with all speech samples, by using the links in the 'Downloads' section at the bottom of the page, and then check all the numbers presented above.


Speech Samples (WAV-files). 
A few independent experts compared the TWELP 600 bps vocoder with the MELPe 600 bps vocoder using the preference method.
All listeners preferred the TWELP vocoder, noting its more natural and less synthetic sound.

You can play and listen to short samples of the source speech, as well as the speech processed by the MELPe 600 bps vocoder and the TWELP 600 bps vocoder, using the links in the table below.

You can also download the complete set of P.50 samples as zip files for all languages simultaneously by using the links in the 'Downloads' section at the bottom of the page.

LanguageSource speechMELPe 600 bpsTWELP 600 bps
American
Arabic
British
Chinese
Danish
Dutch
Finnich
French
German
Greek
Hindi
Hungarian
Italian
Japanese
Norwegian
Polish
Portuguese
Russian
Spanish
Swedish

Superiority In Quality Of The Non-speech Signals. In contrast to other LBR vocoders (MELPe, AMBE+2, etc.), TWELP vocoders provide high quality of non-speech signals, including police, ambulance, fire sirens, etc. This feature in conjunction with high quality natural human-sounding of voice makes TWELP vocoders well suitable for replacement of analog radio by digital radio and also for other applications where high quality transmitting of non-speech signals is relevant along with high quality transmitting of speech signals.

Source signalMELPe 600 bpsTWELP 600 bps
High Robustness To Acoustic Noise. In contrast to other LBR vocoders, TWELP vocoders are highly robust to acoustic noise due to a reliable pitch estimation method and other features of TWELP technology.
 
The diagram demonstrates a significant difference in speech intelligibility for noisy speech (SNR = 10 dB) between the TWELP 600 and MELPe 600 vocoders, based on the ESTOI metric. Exact numbers are shown in the table below.
LanguageTWELP 600 bpsMELPe 600 bps
American 63.51 58.47
Arabic 63.47 59.83
British 61.32 56.73
Chinese 66.61 60.42
Danish 63.88 57.98
Dutch 61.77 56.63
Finnich 59.25 55.25
French 65.31 60.07
German 62.24 56.79
Greek 65.56 59.48
Hindi 60.82 55.59
Hungarian 63.85 59.22
Italian 61.39 57.39
Japanese 66.21 61.64
Norwegian 67.13 61.09
Polish 64.72 59.62
Portuguese 63.59 58.66
Russian 62.02 57.10
Spanish 64.80 59.90
Swedish 57.39 53.69
Average63.2458.28

A difference is on average 4.96 %
Additionally, the TWELP vocoder features an NCSE (Noise Cancellation Speech Enhancement) preprocessor, which cleans the input speech signal from noise and enhances speech quality. 
Below, you can listen to a short fragment of heavily noisy English speech after passing through MELPe and TWELP vocoders, with NPP (Noise Pre-Processor) and NCSE disabled and enabled, respectively.
NPP/NCSEInput speech (SNR=10dB)MELPe 600 bpsTWELP 600 bps
Disabled
Enabled
The NCSE integrated into the TWELP vocoder is described in more detail on the webpage for our standalone product, 'NCSE-AGC Preprocessor'.
High Robustness To The Channel Errors. 
The TWELP technology offers highly efficient speech compression by eliminating redundancy while preserving excellent quality and intelligibility. To enhance robustness against transmission errors, we provide specialized versions called TWELP Robust.
These vocoders are based on an effective Joint Source-Channel Coding approach. Each vocoder is equipped with a custom-designed FEC, tailored to its specific characteristics and operational conditions.
TWELP Robust vocoders provide high speech quality simultaneously in noisy channel as well as in noiseless channel. FEC can operate with "soft decisions" as well as with "hard decisions" from a modem. "Soft decisions" mode provides much better robustness in comparison with the "hard decisions" mode.
For all users of our non-robust vocoder versions, we offer the following recommendations.
 
The diagram below illustrates the sensitivity of bits at the output of the vocoder to communication channel errors.
Essentially, the diagram shows by what percentage speech quality is reduced when a specific bit is distorted. The first bits in order cause catastrophic distortions, while the latter bits have significantly less impact on quality.
 
We strongly recommend using FEC (Forward Error Correction) with unequal protection of the bits in strong accordance with their sensitivity to errors and utilizing 'Soft Decisions' decoding. This will provide the highest robustness of the vocoder against errors in the channel.

 

Additional Functionalities. The following additional functionalities are developed by DSPINI and integrated into TWELP vocoders:

  • Noise Cancellation Speech Enhancement (NCSE)
  • Automatic Gain Control (AGC),
  • Voice Activity Detector (VAD),
  • Discontinuous Transmission (DTX),
  • Tone Detection/Generation (Single tones and Dual tones). The tones are transmitted by the vocoder facilities.

Note: 
The Tone Detector/Generator functionalities are not integrated into the code by default but can be added free of charge upon request.

Each functionality has unique features, performance and characteristics, providing significant superiority over any well-known implementations on the market.

Technical Characteristics And Resource Requirements:

Technical characteristics
Bit Rate
(bps)		AlgorithmFrame size
(ms)		Algorithmic delay
(including frame size)
(ms)		Sampling rate
(kHz)		Signal formatBit stream format
600 TWELP 80 100 8 Linear
16-bit
PCM		 48
Additional functionalities
NameFunctionalityTechnical characteristics
NameValue
AGC Automatic Gain 
Control		 Control range: 0 ... +42 dB
NCSE Noise Canceller -
Speech Enhancer		 SNR increasing  20 dB
Speech quality
improvement		 > 0.1 PESQ
Tone
Detector		 Single/Dual tones 
detection		 In accordance with international standards
Tone
Generator		 Single/Dual tones 
generation		 Special generator, kept continuity of signal 
(phase and amplitude of signal of previous frame)
DTX Discontinuous 
Transmission		 Reduces bit rate down to 110 bps in pauses
between active speech regions
VAD Voice Activity 
Detection		 High reliability even with pink noise at an SNR < 0 dB.
CNG Comfort Noise 
Generation		 Type of noise "white"
Level - 60 dB
The NCSE and AGC integrated into the TWELP vocoder are described in more detail on the webpage for our standalone product, 'NCSE-AGC Preprocessor'.

Resources for ARM Cortex-M4 platform
ModuleMIPS*
peak		Memory (KBytes)
ProgramData
ConstantsChannelHeapStack
Voice Encoder 53 42 201 4.7 5.7 1.0
NCSE 6.0
AGC 0.5
Voice Decoder 19
Voice Encoder +
Voice Decoder		 72
Total 78.5

Resources for TI's C64 DSP platform
ModuleMIPS*
peak		Memory (KBytes)
ProgramData
ConstantsChannelHeapStack
Voice Encoder 19 80 201 4.7 5.7 1.0
NCSE 2.6
AGC 0.3
Voice Decoder 5.3
Voice Encoder +
Voice Decoder		 24.3
Total 27.2

Resources (estimated) for TI's C55 DSP platform
ModuleMIPS*
peak		Memory (KBytes)
ProgramData
ConstantsChannelHeapStack
Voice Encoder 32 26 201 4.7 5.7 1.0
NCSE 6.2
AGC 0.4
Voice Decoder 13
Voice Encoder +
Voice Decoder		 45
Total 51.6

* DSPINI continues optimization of the TWELP algorithm and code in order to minimize computational complexity of the vocoder.

Software Integrity and Security. DSPINI guarantees the ABSOLUTE integrity of its software, free from any undocumented features, undeclared capabilities, or hidden functions. Our customers can be assured that none of our software/code contains any secret features or functionalities concealed from the user. If necessary, we are ready to provide the source code of our software products for appropriate certification.
Moreover, our software is available in source code form—you simply need to purchase the appropriate license to use it.

Guarantee And Support.  DSPINI guarantees a quality and accordance of all technical characteristics of the product to requirement of current specifications. Testing and other method of quality control are used for guarantee support.

Any Platforms.  DSPINI can port this vocoder software into any other DSP, RISC or general- purposes platform inshort time: 1-2 months.

Licensing Terms.  To use the vocoder, customer should obtain a license from DSPINI only.

Customization.  The vocoder can be customized under any specific requirements- other bit rate, frame size, any other robustness to channel errors, etc. Please contact with us for details.

Prospects.  DSPINI is impoving and developing continuously a set of new vocoders with range from 300 bps up to 9600 bps, based on TWELP technology.

Related Software.  This vocoder may be effectively used in a bundle with other DSPINI's products:

  • Linear and acoustic echo cancellers,
  • Multichannel noise cancellers (including two-microphone adaptive array),
  • Wired or radiomodems for any types of channels and bitrates,
  • Other products.

Downloads: