Bing Speech to Text API - 在 c# 中通过 websocket 进行通信

Posted

技术标签:

【中文标题】Bing Speech to Text API - 在 c# 中通过 websocket 进行通信【英文标题】:Bing Speech to Text API - Communicate via websocket in c# 【发布时间】:2017-08-03 19:25:57 【问题描述】:

我正在尝试通过 WebSockets 让 Bing Speech API 在 C# 中工作。我查看了 javascript here 中的实现并一直遵循协议说明 here,但我遇到了一个完整的砖墙。我不能使用现有的 C# 服务,因为我在 Linux 容器中运行,所以我需要使用 .net Core 上的实现。烦人的是,现有的服务是闭源的!

我可以成功连接到网络套接字,但我无法让服务器响应我的连接。我希望收到来自服务器的turn.start 文本消息,但是一旦我发送了几个字节的音频文件,我就会从服务器启动。我知道音频文件的格式正确,因为我直接从 C# 服务示例here 获得它。

我觉得我已经用尽了这里的选项。我现在唯一能想到的就是我没有正确发送音频块。目前,我只是以连续的 4096 字节发送音频文件。我知道第一条音频消息包含只有 36 个字节的 RIFF 标头,然后我只是将它与下一个 (4096-36) 字节一起发送。

这是我的完整代码。您应该能够将其作为 .net 核心或 .net 框架控制台应用程序运行,并且需要一个音频文件和一个 API 密钥。

using Newtonsoft.Json;
using System;
using System.Collections.Generic;
using System.IO;
using System.Linq;
using System.Net.Http;
using System.Net.WebSockets;
using System.Text;
using System.Threading;
using System.Threading.Tasks;

namespace ConsoleApp3

    class Program
    
        static void Main(string[] args)
        
            Task.Run(async () =>
            
                var bingService = new BingSpeechToTextService();
                var audioFilePath = @"FILEPATH GOES HERE";
                var authenticationKey = @"BING AUTHENTICATION KEY GOES HERE";
                await bingService.RegisterJob(audioFilePath, authenticationKey);
            ).Wait();
        
    

    public class BingSpeechToTextService
    
        /* #region Private Static Methods */
        private static async Task Receiving(ClientWebSocket client)
        
            var buffer = new byte[128];

            while (true)
            

                var result = await client.ReceiveAsync(new ArraySegment<byte>(buffer), CancellationToken.None);

                var res = Encoding.UTF8.GetString(buffer, 0, result.Count);

                if (result.MessageType == WebSocketMessageType.Text)
                
                    Console.WriteLine(Encoding.UTF8.GetString(buffer, 0, result.Count));
                
                else if (result.MessageType == WebSocketMessageType.Close)
                
                    Console.WriteLine($"Closing ... reason client.CloseStatusDescription");
                    var description = client.CloseStatusDescription;
                    //await client.CloseOutputAsync(WebSocketCloseStatus.NormalClosure, "", CancellationToken.None);
                    break;
                
                else
                
                    Console.WriteLine("Other result");
                
            
        
        /* #endregion Private Static Methods */
        /* #region Public Static Methods */
        public static UInt16 ReverseBytes(UInt16 value)
        
            return (UInt16)((value & 0xFFU) << 8 | (value & 0xFF00U) >> 8);
        
        /* #endregion Public Static Methods */
        /* #region Interface: 'Unscrypt.Bing.SpeechToText.Client.Api.IBingSpeechToTextJobService' Methods */
        public async Task<int?> RegisterJob(string audioFilePath, string authenticationKeyStr)
        
            var authenticationKey = new BingSocketAuthentication(authenticationKeyStr);
            var token = authenticationKey.GetAccessToken();
            /* #region Connect web socket */
            var cws = new ClientWebSocket();
            var connectionId = Guid.NewGuid().ToString("N");
            var lang = "en-US";
            cws.Options.SetRequestHeader("X-ConnectionId", connectionId);
            cws.Options.SetRequestHeader("Authorization", "Bearer " + token);
            Console.WriteLine("Connecting to web socket.");
            var url = $"wss://speech.platform.bing.com/speech/recognition/interactive/cognitiveservices/v1?format=simple&language=lang";
            await cws.ConnectAsync(new Uri(url), new CancellationToken());
            Console.WriteLine("Connected.");
            /* #endregion*/

            /* #region Receiving */
            var receiving = Receiving(cws);
            /* #endregion*/

            /* #region Sending */
            var sending = Task.Run(async () =>
            


                /* #region Send speech.config */
                dynamic speechConfig =
            new
            
                context = new
                
                    system = new
                    
                        version = "1.0.00000"
                    ,
                    os = new
                    
                        platform = "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (Khtml, like Gecko) Chrome/59.0.3071.115 Safari/537.36",
                        name = "Browser",
                        version = ""
                    ,
                    device = new
                    
                        manufacturer = "SpeechSample",
                        model = "SpeechSample",
                        version = "1.0.00000"
                    
                
            ;

                var requestId = Guid.NewGuid().ToString("N");
                var speechConfigJson = JsonConvert.SerializeObject(speechConfig, Formatting.None);
                StringBuilder outputBuilder = new StringBuilder();
                outputBuilder.Append("path:speech.config\r\n"); //Should this be \r\n
                outputBuilder.Append($"x-timestamp:DateTime.UtcNow.ToString("yyyy-MM-ddTHH:mm:ss.fffK")\r\n");
                outputBuilder.Append($"content-type:application/json\r\n");
                outputBuilder.Append("\r\n\r\n");
                outputBuilder.Append(speechConfigJson);
                var strh = outputBuilder.ToString();

                var encoded = Encoding.UTF8.GetBytes(outputBuilder.ToString());
                var buffer = new ArraySegment<byte>(encoded, 0, encoded.Length);

                if (cws.State != WebSocketState.Open) return;
                Console.WriteLine("Sending speech.config");
                await cws.SendAsync(buffer, WebSocketMessageType.Text, true, new CancellationToken());
                Console.WriteLine("Sent.");
                /* #endregion*/

                /* #region Send audio parts.  */
                var fileInfo = new FileInfo(audioFilePath);
                var streamReader = fileInfo.OpenRead();

                for (int cursor = 0; cursor < fileInfo.Length; cursor++)
                

                    outputBuilder.Clear();
                    outputBuilder.Append("path:audio\r\n");
                    outputBuilder.Append($"x-requestid:requestId\r\n");
                    outputBuilder.Append($"x-timestamp:DateTime.UtcNow.ToString("yyyy-MM-ddTHH:mm:ss.fffK")\r\n");
                    outputBuilder.Append($"content-type:audio/x-wav");

                    var headerBytes = Encoding.ASCII.GetBytes(outputBuilder.ToString());
                    var headerbuffer = new ArraySegment<byte>(headerBytes, 0, headerBytes.Length);
                    var str = "0x" + (headerBytes.Length).ToString("X");
                    var headerHeadBytes = BitConverter.GetBytes((UInt16)headerBytes.Length);
                    var isBigEndian = !BitConverter.IsLittleEndian;
                    var headerHead = !isBigEndian ? new byte[]  headerHeadBytes[1], headerHeadBytes[0]  : new byte[]  headerHeadBytes[0], headerHeadBytes[1] ;

                    //Audio should be pcm 16kHz, 16bps mono
                    var byteLen = 8192 - headerBytes.Length - 2;
                    var fbuff = new byte[byteLen];
                    streamReader.Read(fbuff, 0, byteLen);

                    var arr = headerHead.Concat(headerBytes).Concat(fbuff).ToArray();
                    var arrSeg = new ArraySegment<byte>(arr, 0, arr.Length);

                    Console.WriteLine($"Sending data from cursor");
                    if (cws.State != WebSocketState.Open) return;
                    cursor += byteLen;
                    var end = cursor >= fileInfo.Length;
                    await cws.SendAsync(arrSeg, WebSocketMessageType.Binary, true, new CancellationToken());
                    Console.WriteLine("Data sent");

                    var dt = Encoding.ASCII.GetString(arr);



                
                await cws.SendAsync(new ArraySegment<byte>(), WebSocketMessageType.Binary, true, new CancellationToken());
                streamReader.Dispose();
                /* #endregion*/

                
                    var startWait = DateTime.UtcNow;
                    while ((DateTime.UtcNow - startWait).TotalSeconds < 30)
                    
                        await Task.Delay(1);
                    
                    if (cws.State != WebSocketState.Open) return;
                
            );
            /* #endregion*/

            /* #region Wait for tasks to complete */
            await Task.WhenAll(sending, receiving);
            if (sending.IsFaulted)
            
                var err = sending.Exception;
                throw err;
            
            if (receiving.IsFaulted)
            
                var err = receiving.Exception;
                throw err;
            
            /* #endregion*/

            return null;

        
        /* #endregion Interface: 'Unscrypt.Bing.SpeechToText.Client.Api.IBingSpeechToTextJobService' Methods */


        public class BingSocketAuthentication
        
            public static readonly string FetchTokenUri = "https://api.cognitive.microsoft.com/sts/v1.0";
            private string subscriptionKey;
            private string token;
            private Timer accessTokenRenewer;

            //Access token expires every 10 minutes. Renew it every 9 minutes.
            private const int RefreshTokenDuration = 9;

            public BingSocketAuthentication(string subscriptionKey)
            
                this.subscriptionKey = subscriptionKey;
                this.token = FetchToken(FetchTokenUri, subscriptionKey).Result;

                // renew the token on set duration.
                accessTokenRenewer = new Timer(new TimerCallback(OnTokenExpiredCallback),
                                               this,
                                               TimeSpan.FromMinutes(RefreshTokenDuration),
                                               TimeSpan.FromMilliseconds(-1));
            

            public string GetAccessToken()
            
                return this.token;
            

            private void RenewAccessToken()
            
                this.token = FetchToken(FetchTokenUri, this.subscriptionKey).Result;
                Console.WriteLine("Renewed token.");
            

            private void OnTokenExpiredCallback(object stateInfo)
            
                try
                
                    RenewAccessToken();
                
                catch (Exception ex)
                
                    Console.WriteLine(string.Format("Failed renewing access token. Details: 0", ex.Message));
                
                finally
                
                    try
                    
                        accessTokenRenewer.Change(TimeSpan.FromMinutes(RefreshTokenDuration), TimeSpan.FromMilliseconds(-1));
                    
                    catch (Exception ex)
                    
                        Console.WriteLine(string.Format("Failed to reschedule the timer to renew access token. Details: 0", ex.Message));
                    
                
            

            private async Task<string> FetchToken(string fetchUri, string subscriptionKey)
            
                using (var client = new HttpClient())
                
                    client.DefaultRequestHeaders.Add("Ocp-Apim-Subscription-Key", subscriptionKey);
                    UriBuilder uriBuilder = new UriBuilder(fetchUri);
                    uriBuilder.Path += "/issueToken";

                    var result = await client.PostAsync(uriBuilder.Uri.AbsoluteUri, null);
                    Console.WriteLine("Token Uri: 0", uriBuilder.Uri.AbsoluteUri);
                    return await result.Content.ReadAsStringAsync();
                
            
        
    

【问题讨论】:

【参考方案1】:

我知道这很简单。

经过几个小时令人沮丧的编码后,我发现了问题所在。我一直忘记发送request id 以及speech.config 电话。

【讨论】:

如果有人感兴趣,这里github上有完整的demo:github.com/maptz/maptz.speechtotext.tool

以上是关于Bing Speech to Text API - 在 c# 中通过 websocket 进行通信的主要内容,如果未能解决你的问题,请参考以下文章

要发送到 Bing Speech to Text API 的最大音频文件长度(持续时间)应该是多少?

Bing Text-to-Speech 可以采用 javascript 变量值并将其转换为语音吗?

如何在 Bot Framework C# 中使用 Bing Speech API

如何从 Google Cloud API Text-to-Speech 获取音素

请求 Google Text-To-Speech API [关闭]

与 Bing Speech API 集成的 Java Web